Increased non-prostate cancer death risk in clinically diagnosed prostate cancer
Pim J. van Leeuwen, Erasmus MC, University Medical Centre, Room NH 227, P.O. Box 2040, 3000 CA Rotterdam, The Netherlands. e-mail: firstname.lastname@example.org
Study Type – Prognosis (case control)
Level of Evidence 3a
What's known on the subject? and What does the study add?
Treatment of advanced PC might put patients at an increased risk of cardiovascular events. Recent studies have suggested that the excess mortality is lower among men who were diagnosed with screen detected PC in comparison to men with clinically diagnosed PC, possibly due to the use of medications for cardiovascular disease and the change to a healthier lifestyle of men with a screen detected PC.
Men with clinically diagnosed PC have an increased risk of death unrelated to PC itself, i.e., the excess mortality is based on an increased risk of dying from other neoplasm and diseases of the circulatory or respiratory system.
- • To assess the cause-specific mortality unrelated to prostate cancer (PC) itself in patients with screen- and clinically diagnosed PC.
PATIENTS AND METHODS
- • The present study was conducted among participants of the European Randomized Study of Screening for Prostate Cancer.
- • Based on consensus of the causes of death committee (CODC), all patients who died from PC were excluded.
- • In the intervention arm, cases were patients with a screen-detected PC, aged 55–74 years, between 1993 and 2001.
- • These cases were matched to two controls in whom no cancer was found after biopsy, and two controls in whom no cancer was suspected after screening. In the control arm, cases were patients with clinically diagnosed PC, aged 55–74 years, between 1993 and 2001. These cases were matched to four controls without PC. Matching was done with respect to date of birth, screening and/or diagnosis. Men were followed up to 31 December 2007.
- • No statistically significant difference in overall mortality between cases and controls in the intervention arm was observed: relative risk (RR) 1.26 (95% confidence interval [CI] 0.96–1.65; P = 0.102) and RR 1.13 (95% CI 0.86–1.47; P = 0.381).
- • In the control arm, the overall mortality was statistically significantly higher in cases relative to controls: RR 1.43 (95% CI 1.03–2.00; P = 0.033).
- • This difference was because of an increased risk of dying from neoplasms and disease of the circulatory or respiratory system among cases: RR 1.61 (95% CI 1.12–2.29; P = 0.009).
- • The present study was limited by the relatively small sample size.
- • Increased mortality unrelated to PC itself was observed in men with clinically diagnosed PC, but not in screen-detected PC.
- • The excess mortality in men with clinically diagnosed PC seems to be as a result of a significantly increased risk of dying from neoplasm and disease of the circulatory or respiratory system.
- • Results have to be studied more thoroughly in further clinical trials.
European Randomized Study of Screening for Prostate Cancer
International Classification of Diseases
causes of death committee
Prostate cancer (PC) has become the most common non-cutaneous diagnosed cancer in men in Europe and the United States . Currently, only a small percentage of men with predominantly localized PC die from a PC-related cause of death. Cardiovascular disease (CVD) is the primary or secondary cause of death in most PC patients .
Recent evidence has suggested that there might be a positive correlation between increased cardiovascular risk and the incidence, progression and treatment for PC [3–8]. Some methods to reduce the risk for CVD seem to be similar to methods to reduce the risk for high-risk PC [9–11]. In addition, recent evidence has suggested that the excess mortality is lower among men who were diagnosed with screen-detected PC in comparison to men with clinically diagnosed PC, possibly as a result of the use of medications for CVD and/or the change to a healthier lifestyle of men with screen-detected PC . The present study was designed to quantify the excess mortality measured in the previous study by van Leeuwen et al. .
Previously we have demonstrated that CVD mortality is not increased among men diagnosed with PC in the Rotterdam section of the European Randomized Study of Screening for Prostate Cancer (ERSPC) compared with the general Dutch population . However, no distinction was made between screen-detected and symptomatically detected patients, and to socio-economic class, which is known to be higher in men participating in the ERSPC in comparison to the general population. Making use of the detailed information on each individual participant in the ERSPC Rotterdam, we compared the incidence of overall mortality and non-prostate cancer causes of death among men without PC and men with screen-detected PC on the one, and unscreened men with symptomatically diagnosed PC on the other hand. The aim of this analysis was to assess whether men with screen-detected and symptomatically diagnosed PC are at an increased risk of death and of which particular causes. The comparison group for men with screen-detected PC consisted of men who were screened for PC but in whom no PC was detected. Unscreened men in the control arm with symptomatically PC were compared with men participating in the control population of the ERSPC Rotterdam who were not diagnosed with PC.
MATERIALS AND METHODS
All men were participants of the Rotterdam section of the ERSPC. All men signed an informed consent before randomization to systematic screening (intervention arm) and usual care (control arm).
Men in the intervention arm were screened with an interval of 4 years by PSA measurement, DRE and TRUS between December 1993 and May 1997. A sextant biopsy was initially offered to men with a PSA level ≥4.0 ng/mL and/or a suspicious finding on DRE and/or TRUS. After May 1997, a biopsy was prompted by a PSA level ≥3.0 ng/mL only. All cancers were classified according to the 1992 TNM classification. Men who were diagnosed with PC were treated by the local urologist. Details of the screening methodology were reviewed by Roobol and Schroder .
The present study was designed as a prospective cohort study including cases and controls where case subjects were men diagnosed with PC (screen-detected or symptomatically) and control subjects were men who were not diagnosed with PC until death or the cut-off date of the 31 December 2007. Matching was performed to ensure equal risks of non-PC death among the case and control subjects.
All males, aged 55–74 years, in the intervention arm of ERSPC Rotterdam diagnosed with a screen-detected PC between 1 January 1995 and 31 December 2004 were eligible as case subjects (case intervention arm) and were actually selected if the tumour was diagnosed at screening and was localized (defined as stage T1C, N0, M0 and serum PSA level <20.0 ng/mL). Cases were excluded if they were determined as PC-related cause of death before the 31 December 2007. These inclusion criteria were established to select only screen-detected PCs. Only T1c cancers were included as T1c PC is the disease diagnosed in a patient at screening without any clinical sign that they have the disease. The date of diagnosis of the case was the index date. The control subjects were randomly selected from the intervention arm, alive at the date of diagnosis of the case and were matched to the case subjects by month and year of birth, self-reported health status (i.e. good, moderate or poor) and month and year of screening. Each time a case subject was included in the present study, we randomly selected four controls; two control subjects in whom no cancer was found after a prostate biopsy and two control subjects who had a normal PSA value (PSA level <3.0 ng/mL) and did not undergo prostate biopsies or were not suspicious of having cancer. If there was no match in respect to the previously mentioned variables, then a control subject was selected who was born 1 month before or after the case subject. If that still did not lead to a match, then a control subject was selected born 2 months after the case subject and then, if necessary, 2 months before the case subject was born, and so forth, until a match was found with a maximum difference of 6 months. The two most optimal groups were chosen as both might be criticisable for a reason. The aim of the present study design was to indentify controls who did not have PC. We selected men with a PSA level <3.0 ng/mL at a screening visit although it is known that there is no PSA value where PC is not detectable , and men with no cancer found at a sextant biopsy although it is known that a sextant prostate biopsy misses a substantial percentage of PC .
These case subjects, aged 55–74 years, were all men randomized to the control arm of the ERSPC, section Rotterdam, who were clinically diagnosed with PC, defined as stage > T1c, N0/N1, M0/M1 between 1 January 1995 and 31 December 2004 (case–control arm). Cases were excluded if they were determined as having died from a PC-related cause before the 31 December 2007. T1c cancers were excluded as T1c PC is the disease diagnosed in a patient at screening without any clinical sign they have the disease. The controls were randomly selected from the control arm, alive at the date of diagnosis of the case, and matched to the case subjects on month and year of diagnosis with respect to month and year of birth, self-reported health status (i.e. good, moderate or poor) and month and year of randomization. Each time a case subject was included in the present study, we randomly selected four controls. Control subjects were men who were randomized to the control arm of the ERSPC, section Rotterdam, alive at the time of diagnosis of the case, and not diagnosed with PC. If there was no match with respect to the previously mentioned variables, the same complementary matching process was performed as described above.
Mortality data of case and control subjects in both the intervention and the control arm that died in the period up to 31 December 2007 were obtained by linking the trial database with the Causes of Death Registry of Statistics Netherlands. Causes of death were based on the national certificates coded according to the International Classification of Diseases (9th revision [ICD9] 1995) and ICD10 (from 1996 onward), and the causes of death grouping was based on the tabulation list for main primary causes of death of Statistics Netherlands . For all case subjects in the study who were known to have died, all available information was gathered and anonymized. Subsequently, cases classified as having died from a PC-related death by the independent causes of death committee (CODC) of the ERSPC Rotterdam were excluded from the study. The CODC reviews all deceased PC cases using predefined flow charts. Patients were determined to have died from PC if they were classified as ‘definitely PC death’, ‘possible PC death’ or as ‘PC intervention-related death’.
For the statistical analyses, the mortality rates were compared between cases and controls using a Poisson regression analysis with an indicator of the study arm as a predictor and the logarithm of the number of person years as an offset term (predictor with a coefficient of one) . For the cases and controls, the time of follow-up was measured from the date of randomization up to date of death or 31 December 2007. P values <0.05 were considered statistically significant. All analyses were performed with the commercially available STATA package version 11 (STATA Corporation, College Station, TX, USA).
A total of 372 cases and 1488 controls participating in the intervention arm of the ERSPC section Rotterdam were included in the present study, and a total of 221 cases and 884 controls participated in the control arm of the ERSPC section Rotterdam. The age at diagnosis and tumour characteristics of the cases are presented in Table 1. The median age at diagnosis differed significantly between the two groups (P < 0.001); cases in the intervention arm were diagnosed at an earlier age.
Table 1. Baseline characteristics at diagnosis
|Total participants included||372||221|| |
|Age (years), median||66.6||68.5||<0.01|
| 55–60||68 (18.3)||19 (8.6)||<0.01|
| 61–65||107 (28.8)||56 (25.3)|| |
| 66–70||115 (30.9)||79 (35.8)|| |
| 71–74||82 (22.0)||67 (30.3)|| |
|PSA at diagnosis (ng/mL), median||4.7||11.6||<0.001|
|Tumour stage|| || || |
| T1c||372 (100.0)||–|| |
| T2||–||141 (63.8)|
| T3||–||76 (34.4)|
| T4||–||4 (1.8)|
|Histological differentiation|| || || |
| Gleason 2–6||293 (78.8)||90 (40.7)||<0.001|
| Gleason 7||70 (18.8)||77 (34.8)|
| Gleason 8–10||9 (2.4)||30 (13.6)|
| Not known or not performed||–||24 (10.9)|
|Primary treatment|| || || |
| Surgery||156 (41.9)||59 (26.7)|| |
| Radiotherapy||131 (35.3)||116 (52.5)|
| Watchful waiting||83 (22.3)||13 (5.9)|
| Hormone||–||30 (13.6)|
| Other||2 (0.5)||3 (1.3)|
|Adjuvant treatment|| || || |
| Hormone||4 (1.1)||28 (12.7)|| |
The median follow-up from the time of diagnosis up to either death or end of follow-up was 8.9 years for the cases. Up to the end of 2007, a total of 82 (22.0%) cases died (26.2 men per 1000 person years). A total of 139 (18.7%) controls died in whom no cancer was found after biopsy, and a total of 155 (20.8%) controls in whom no cancer was suspected (Table 2). This resulted in an all-cause mortality that was not significantly different between cases and controls: relative risk (RR) 1.26 (95% CI 0.96–1.65; P = 0.102) for the cases relative to the controls in whom no cancer was found after biopsy; and RR 1.13 (95% CI 0.86–1.47; P = 0.381) for the cases relative to the controls in whom no cancer was suspected. In addition, no significant difference was found between the cases and all controls in the intervention arm; RR 1.19 (95% CI 0.93–1.52; P = 0.168).
Table 2. Causes of death in the intervention arm stratified by study group
|Total included||372 (100)||1488 (100)|| ||744 (100)|| ||744 (100)|| |
|Total no. of deaths||82 (22.0)||294 (19.8)||1.19 (0.93–1.52)||139 (18.7)||1.26 (0.96–1.65)||155 (20.8)||1.13 (0.86–1.47)|
|Total CVD deaths||26 (7.0)||100 (6.7)||1.11 (0.72–1.70)||47 (6.3)||1.18 (0.73–1.90)||53 (7.1)||1.04 (0.65–1.67)|
|Neoplasms||40 (10.7)||118 (7.9)|| ||54 (7.2)|| ||64 (8.6)|| |
|Respiratory system||1 (0.2)||20 (1.3)|| ||7 (0.9)|| ||13 (0.5)|| |
|Digestive system||2 (0.5)||7 (0.5)|| ||4 (0.5)|| ||3 (0.4)|| |
|Endocrine and metabolic||1 (0.3)||10 (0.7)|| ||5 (0.7)|| ||5 (0.7)|| |
|Abnormal clinical or laboratory findings||3 (0.8)||11 (0.7)|| ||7 (0.9)|| ||4 (0.5)|| |
With respect to CVD, a total of 26 (7.0%) cases died from CVD (8.3 men per 1000 person years) and a total of 47 (6.3%) controls in whom no cancer was found after biopsy, and a total of 53 (7.1%) controls in whom no cancer was suspected. Consequently, no statistically significant difference in CVD was observed between the cases and controls; RR 1.18 (95% CI 0.73–1.90; P = 0.503) for the cases relative to the controls in whom no cancer was found after biopsy; and RR 1.04 (95% CI 0.65–1.67; P = 0.854) for the cases relative to the controls in whom no cancer was suspected. The RR on CVD was 1.11 (95% CI 0.72–1.70; P = 0.643) for all cases relative to all the controls.
The six most frequent causes of death among the cases and controls in the intervention arm are presented in Table 2. Neoplasms were the most frequent cause of death, followed by CVD. Despite all PC-related deaths based on the consensus of the CODC were excluded, in three cases the primary cause of death was PC (3.6% of all deaths). The mortality from neoplasms, CVD and diseases of the respiratory system together was not statistically significant different in the cases relative to the controls; RR 1.20 (95% CI 0.92–1.58; P = 0.180).
For the cases, the median follow-up from time of diagnosis up to either death or end of follow-up was 6.1 years. A total of 47 (21.3%) cases (34.2 men per 1000 person years) and 134 (15.2%) controls died (Table 3). This resulted in an all-cause mortality that was statistically significantly different between cases and controls: RR 1.43 (95% CI 1.03–2.00; P = 0.033). CVD was not statistically significantly different between the cases and controls (Table 3). A total of 11 (5.0%) cases (8.0 men per 1000 person years) and 30 (3.4%) controls died from CVD; RR 1.50 (95% CI 0.75–2.99; P = 0.250) for cases relative to the controls.
Table 3. Causes of death in the control arm stratified by study group
|Total included||221 (100)||884 (100)|| |
|Total no. of deaths||47 (21.3)||134 (15.2)||1.43 (1.03–2.00)|
|Total CVD deaths||11 (5.0)||30 (3.4)||1.50 (0.75–2.99)|
|Neoplasms||26 (11.8)||62 (7.0)|| |
|Respiratory system||5 (2.3)||15 (1.7)|| |
|Digestive system||2 (0.9)||4 (0.4)|| |
|Endocrine and metabolic||1 (0.4)||2 (0.2)|| |
|Abnormal clinical or laboratory findings||0 (0.0)||6 (0.7)|| |
The six most frequent causes of death among the cases and controls in the control arm are presented in Table 3. Neoplasms were the most frequent cause of death followed by CVD and diseases of the respiratory system. Based on data of the Causes of Death Registry of Statistics Netherlands, a total of six death were recorded as primary cause PC, although all PC-related deaths based on the consensus of the CODC were excluded (12.8% of all deaths). The mortality from neoplasms, CVD and diseases of the respiratory system together was statistically significantly higher among the cases relative to the controls; RR 1.61 (95% CI 1.12–2.29; P = 0.009).
In the present study, men with clinically diagnosed PC were at an increased risk of dying from a cause unrelated to PC itself compared with men without PC, RR: 1.43 (95% CI 1.03–2.00; P = 0.033) if all PC deaths based on the consensus of the CODC were excluded. This difference was based on an increased risk of dying from neoplasms and diseases of the circulatory and respiratory system, together: RR 1.61 (95% CI 1.12–2.29; P = 0.009). Results are limited by a relatively small sample size. The observations are in line with previous findings of the ERSPC section Rotterdam in which a secondary mortality analysis showed that the excess mortality was higher than the disease-specific mortality in the control arm of the ERSPC Rotterdam. This previous study suggested a possible underestimation of deaths from PC by the CODC in the control arm of the study, or an additional disease-related mortality that was measured by an excess mortality analysis but not with a disease-specific mortality analysis . The present study confirms that both these suggestions have a contribution to the observed excess mortality in the control arm of the study.
All cases in the present study that were determined using the CODC to have died from a PC-related cause of death were excluded from the study. For this reason, in theory, the overall mortality in cases and controls was expected to be similar. However, this was not the case in the control arm of the study. Based on data of the death certificates, three cases with a screen-detected PC (0.81% of all cases) and six cases with a clinically diagnosed PC (2.71% of all cases) were recorded with PC being the primary cause of death. This observation, despite being based on small numbers, suggests that the accuracy of the cause of death determination is higher in screen-detected than in clinically diagnosed PC. As suggested previously, it might be possible that the reviewers have underestimated the PC mortality by being too cautious in stating that a patient has died from PC in the present study and that the information to make a judgement is more limited in the control arm of the study. Furthermore, it may be that as a result of the study protocol, more information could be gathered from patients diagnosed with PC in the intervention arm resulting in a higher accuracy of the death certificates.
What are the reasons why men with PC have an increased risk of dying from causes unrelated to PC itself. First, recently, Van Hemelrijck et al.  observed an increased risk of non-fatal and fatal CVD in patients with PC receiving curative treatment, surveillance or especially androgen-deprivation therapy (ADT). Findings that are in line with previous studies; a systematic review of four studies showed that men who underwent ADT had a significantly increased risk of CVD (summary hazard ratio, 1.17; 95% CI, 1.07–1.29) . ADT is also associated with osteoporosis and fractures, diabetes and with an increased risk of colorectal cancer [21–23]. Given the widespread use of ADT to treat especially locally advanced PC, treatment with ADT might result in an increased risk of death in patients with clinically diagnosed PC in particular. In the present study, 26.6% of the cases in the control arm received primary and/or adjuvant treatment with ADT in contrast to 1.1% of the cases in the intervention arm (Table 1). Second, the median age at diagnosis in men participating in the control arm was higher than men in the screening arm. Consequently, the increased risk of dying from causes unrelated to PC itself might be associated with the increasing age at diagnosis. However, no effect of age at diagnosis could be measured among men diagnosed with PC in the screening arm.
The risk of excess mortality in screen-detected PC was suggested to be lower as a consequence of changes in medical regimen, medication and lifestyle unrelated to PC itself . It is known that the incidence of clinically insignificant PC is large among men participating in a screening programme . Nevertheless, a large percentage of these men obtain invasive treatment with curative intent. Consequently, after a PC diagnosis, the first contact with the urologist and/or radiation oncologist often involves screening of patients' general health including the measures of the vital signs that lead to other interventions that might be associated with the decrease in non-PC-specific mortality. It has been proven that the use of these medications is increased in men and women diagnosed with PC or breast cancer subsequent to screening [25–28]. Furthermore, the change in medical treatments is studied in detail in a previous study among 180 men with screen-detected PC. This study showed among men diagnosed with screen-detected PC a significant change in medical treatments and prescriptions unrelated to PC itself subsequent to their diagnosis . In total 72% of the men had a change in medical regimen after diagnosis, 61% had a change in medication and 29% received a new medical diagnosis. In total, 24 (14%) had a multiple gated acquisition scan, treadmill or persantine thallium test, and 23 (13%) had another treatment performed, most frequently a cardiac catheterization. Among the men with changes to their medical regimen 36 (59%) began a new or different medication, which in 67% was a beta-blocker . Meta-analyses including 23 randomized controlled trials showed a statistically significantly lower CVD and all-cause mortality (relative risk 0.76, 95% CI: 0.68, 0.84) in the groups treated with beta-blockers than in the control groups after a mean of 1 year . Other frequently new started medication was statins. A meta-analyses including nine studies showed that the use of statins for secondary prevention in elderly patients with documented coronary heart disease reduced all-cause mortality by 22% and reduced coronary heart disease mortality by 30% with median use of 3 years [30–32]. Also, relatively frequently new started medications were inhaled corticosteroid and long-acting beta-2 agonists for the treatment a chronic obstructive pulmonary disease. Using the combination of inhaled corticosteroid and long-acting beta-2 agonist therapy for more than 6 months has been shown to be associated with a 20% reduced total mortality . Although these medication effects are shown in large randomized controlled trials, we might suggest that they have at least contributed to the observed effect among men with screen-detected cancers in the present study.
As there is emerging evidence that lifestyle factors can alter the rate of progression of indolent PC, many men frequently take on the responsibility of improving their general health by making lifestyle and dietary changes [34,35]. A small randomized clinical trial showed that intervention participants had significantly improved their lifestyle compared with controls at 12 months . Recommendations given to prevent progression of indolent PC include for example, choosing nutritious foods, focusing on fruits and vegetables, drinking green tea, eating omega-3 rich foods, choosing healthy fats, manage stress and taking supplements. Consequently, these recommendations might also have an effect on the non-PC related mortality, incidence and mortality rates of other neoplasms. Urologists might, therefore, be aware of positive effects of life-style modification and might give more attention to co-morbidities. Further, urologists might develop programmes in conjunction with for example cardiologists to counsel asymptomatic men before initiating early PC treatment, in order to identify those who are at an increased risk of non-PC related mortality. However, the results presented in the present study have a low level of evidence, which should not automatically support the changes in medical treatment in men with screen-detected PC. Therefore, this issue should be studied more thoroughly in further analyses or other clinical trials. For example, the effect of the change in medical treatments might be studied in detail to assess the specific effect of the various medications on the outcomes in men with screen-detected PC.
There are several limitations in the present study. The number of deaths was relatively small. Consequently the CIs for the estimates of the cause-specific death groups were rather wide. No data were available on smoking, body mass index and serum lipid levels, although these are strong risk factors for CVD. Although the present study reached the optimal match between cases and controls that ensured an equal risk of non-PC death, it remains possible that especially the cases in the control arm were less healthy as part of their late diagnosed clinically significant PC (possibly these men refused the health system in general). In addition, the potential role of the use of medication that decreases the risk of CVD was not assessed in the present study. Finally, data from this study after longer follow-up are needed to confirm current observations. The strength of the present study includes the validity of the control cases. In general, control cases are obtained from the general population although PC patients often have been shown to be from a selection of men considering general health and social economic class. The present study also has strong aspects. Participants were part of a randomized controlled trial and were followed by regular matching with national cancer registries. Cases and controls were part of a homogeneous study population.
In conclusion, men with clinically diagnosed PC have an increased risk of death unrelated to PC itself. This excess mortality was probably as a result of a significantly increased risk of dying from a neoplasm and disease of the circulatory or respiratory system. No increased risk in all cause-mortality unrelated to PC itself was observed among men diagnosed with PC subsequent to screening. Many effects might have influenced these observations, however, the relatively increased use of ADT in clinically diagnosed PC and the change of medical regimes and medication among men with screen detected PC may have had the most influence. These results should be studied more thoroughly in further clinical trials. The present study is limited by the relatively small sample size and small number of events. Finally, if changes in the medical regimens really do affect all-cause mortality in men with PC, uro-oncologists should look carefully at the management of abnormal parameters of the circulatory and respiratory system, and should encourage PC patients to make lifestyle modifications.
The ERSPC is supported by grants from the Dutch Cancer Society (KWF 94-869, 98-1657, 2002-277, 2006-3518), The Netherlands Organization for Health Research and Development (002822820, 22000106, 50-50110-98-311), 6th Framework Program of the EU: P-Mark: LSHC-CT-2004-503011, Beckman Coulter Hybritech Inc and of Europe against Cancer (SOC 95 35109, SOC 96 201869 05F02, SOC 97 201329, SOC 98 32241). The ERSPC received Erasmus MC and Ministry of Health institutional review board approval.
CONFLICT OF INTEREST
Franz H. Schröder is an advisor to GlaxoSmithKline, Ferring, and Schering. Monique J. Roobol is an advisor for GlaxoSmithKline, Beckman and Genprobe.