Localised prostate cancer in elderly men aged 80–89 years, findings from a population-based registry
Abstract
Objectives
To investigate the rate of prostate cancer-specific mortality (PCSM) and disease characteristics in patients diagnosed with localised prostate cancer at age 80–89 years in comparison with men diagnosed at age 70–79 years.
Patients and Methods
This is a retrospective study of data from the South Australian Prostate Cancer Clinical Outcomes Collaborative (SA-PCCOC). Included were men diagnosed between 2005 and 2014, aged ≥70 years with no evidence of metastatic disease at presentation. Propensity score matching and competing risk Fine and Grey regression were used to assess the chance of treatment (curative vs non-curative) and treatment effect on PCSM.
Results
Of the 1 951 eligible patients, 1 428 (76%) were aged 70–79 years and 460 (24%) were aged 80–89 years at diagnosis, with a median (interquartile range) age of 74 (72–76) and 83 (81–85) years, respectively. The 80–89 years group had higher Gleason scores and Prostate Specific Antigen (PSA) values (all P < 0.001) in comparison with the younger group. The 80–89 years group were less likely to be treated with curative treatment (odds ratio 0.12, 95% confidence interval 0.09–0.16; P < 0.001). The proportion of deaths attributable to prostate cancer was similar in both groups: 73 of 263 deaths (28%) in the 80–89 years group vs 97 of 310 deaths (31%) in the 70–79 years group. The risk of PCSM in individuals treated with curative intent was reduced in both groups.
Conclusions
The proportion of prostate cancer deaths was similar in both groups. These findings support carefully selected individualised management of elderly patients diagnosed with localised prostate cancer.
Introduction
The proportion of the population aged ≥80 years is rising globally, with continuing improvement in life-expectancy in this group. Currently, there are no published data available on patients who are diagnosed with localised prostate cancer at age ≥80 years.
Prostate cancer is the second most common cancer in men and one of the leading causes of cancer-related mortality 1. Both prostate cancer prevalence 2 and incidence 3 increase with age, up to the age of 80–85 years 4. Compared with younger men, elderly men are more likely to have advanced stage, higher grade, and clinically significant prostate cancer 5. Elderly patients with localised prostate cancer are less likely to receive treatment with curative intent 6, 7. There is an emerging body of evidence to support active treatment in selected elderly men with localised prostate cancer 8, 9.
The objectives of the present study were to investigate the rate of prostate cancer-specific mortality (PCSM) and disease characteristics in patients diagnosed with localised prostate cancer at age 80–89 years, in comparison with younger patients (70–79 years), and to describe patient's and disease characteristics, management approaches, causes of death (prostate cancer death vs non-prostate cancer death), and prostate cancer-specific survival in the study population.
Patients and Methods
Data Collection
The South Australian Prostate Cancer Clinical Outcomes Collaborative (SA-PCCOC) is a disease-specific, prospective, third-party collected, longitudinal, observational registry in the Australian state of South Australia (SA). The registry currently captures ~90% of all new histologically confirmed diagnoses of prostate cancer in SA. It began in 1998, and is the longest running prostate cancer registry in Australia and the Southern Hemisphere.
Men with prostate cancer were identified using the SA-PCCOC registry. The date of database lock was 02 December 2016. Inclusion criteria were: age at diagnosis of 70–89 years, and date of diagnosis between 2005 and 2014 inclusive. Men with metastatic disease at presentation were excluded. Metastatic disease was identified through case-note records, and imaging results (whole body bone scan, CT, or MRI) within 45 days of diagnosis. Patients with inconsistent diagnosis and treatment dates, or incomplete date and cause of death information were also excluded. Cause of death and date of death were retrieved through linkage to the SA Births, Deaths and Marriages Registry.
Men were divided into two cohorts based on age at diagnosis: 70–79 and 80–89 years. Risk groups were defined according to the National Comprehensive Cancer Network (NCCN) classification. Risk assessment categories were: low risk (stage cT1–T2a and PSA level ≤10 ng/mL and Gleason score ≤6), intermediate risk (stage cT2b/c and/or Gleason score 7 and/or PSA level 10–≤20 ng/mL), and high-risk (cT stage ≥3a or PSA level >20 ng/mL or Gleason score 8–10) 10. When an individual's disease characteristic data were incomplete, that person was assumed to be in the lowest risk group given the available data. Curative-intent treatments included radical prostatectomy (RP), brachytherapy, and external beam radiotherapy. Non-curative treatments included hormonal therapies and watchful waiting.
Statistical Methods
An a priori list of risk factors related to age at diagnosis were determined to be hospital setting (care received in public vs private hospital), co-morbidities, Gleason score, T-stage, PSA level, NCCN risk-group, and year of diagnosis. Differences between age groups in these factors were assessed using linear-by-linear trend tests with age divided into 5-year groups.
The association between age and treatment type (curative vs non-curative), conditional on the NCCN risk-group and comorbidity indices was assessed using binomial logistic regression. Risk group and the pairwise interaction with age were included as covariates. In addition, a propensity score matched conditional binomial logistic regression was used to estimate the difference in log odds (i.e. odds ratio [OR]) of receiving curative treatment between octogenarians and men aged 70–79 years. The NCCN risk-group, C3 index 11, Charlson Comorbidity Index, and year of diagnosis were used to predict age group (70–79 vs 80–89 years). A 1:1 matching was employed with a caliper set at 0.1 of the standard deviation (sd) of the logit.
Survival duration was calculated as time from date of diagnosis until date of death or date of database lock. For overall survival (OS) death from any cause was considered an event, and men alive at last contact were censored. In the competing risk analyses death from prostate cancer was considered the event of interest. Follow-up was defined as the median survival in which alive-at-last-contact was considered an event, and deaths were censored. Kaplan–Meier methodology was used to estimate median follow-up, and 5- and 10-year OS rates. A covariate-unadjusted Cox regression was used to determine the difference in hazard between the two age groups for OS.
Associations between PCSM and risk factors of interest, including age (in 5-year age groups: 70–74, 75–79, 80–84, 85–89 years), were assessed using univariate Fine and Gray competing risk models 12. We report prostate cancer-specific sub-distribution hazard ratios (sHRs) and present the cumulative incidence curves.
To estimate the efficacy of curative treatment, Fine and Gray competing risk analyses were performed within age groups (70–79 and 80–89 years). A 2:1 ratio of propensity score matching was performed with the same caliper rule as above, with age (continuous), Gleason score, PSA level, and T-stage as additional matching variables.
In all analyses missing covariate data was imputed with cohort means except for NCCN risk stratification as noted above. A complete case sensitivity analysis was performed excluding individuals missing one of the three NCCN risk stratification variables (T-stage, PSA level or Gleason score at diagnosis). The level of significance was taken at the 0.05 threshold for all analyses. Analyses were performed using R (v3.4.1) with the cmprsk competing risk and nonrandom propensity score matching packages.
Results
At the time of analysis the SA-PCCOC database contained 9 238 prostate cancer cases of whom 2 303 were aged 70–89 years at diagnosis, with a diagnosis between 2005 and 2014. Of these, 415 were excluded due to: the presence of metastatic disease at diagnosis (n = 150), locally advanced disease at diagnosis (n = 148), NCCN very high risk (Gleason score primary 5; n = 71), missing diagnostic data (n = 35), or inconsistent dates of treatment/death (n = 11). Finally, 1 888 cases were available for analysis: 1 428 (76%) aged 70–79 years and 460 (24%) aged 80–89 years at diagnosis (Fig. 1), with a median (interquartile range) age of 74 (72–76) and 83 (81–85) years, respectively.

Men aged 80–89 years at diagnosis were more likely to be treated in the public sector, have more co-morbidity, and more frequently diagnosed with high-risk disease (all P < 0.001). At diagnosis, they had higher Gleason scores (P < 0.001) and PSA values (P < 0.001) than younger men (Table 1). The T-stage was not significantly different between the two groups (P = 0.17).
Age group | ||||
---|---|---|---|---|
Variable | 70–79 years | 80–89 years | All | P |
Number of patients | 1 428 | 460 | 1 888 | |
Hospital type, n (%) | ||||
Public | 931 (65) | 365 (79) | 1 296 (69) | <0.001 |
Private | 391 (27) | 82 (18) | 473 (25) | |
Not reported | 106 (7) | 13 (3) | 119 (6) | |
C3 index | ||||
Mean (sd) | 2.1 (2.5) | 3.3 (2.9) | 2.4 (2.7) | <0.001 |
Not reported, n (%) | 277 (19) | 66 (14) | 343 (18) | |
Charlson Comorbidity Index | ||||
Mean (sd) | 1.3 (1.8) | 1.9 (2.2) | 1.4 (1.9) | <0.001 |
Not reported, n (%) | 277 (19) | 66 (14) | 343 (18) | |
NCCN risk group, n (%) | ||||
Low | 417 (29) | 84 (18) | 501 (27) | <0.001 |
Intermediate | 721 (50) | 175 (38) | 896 (47) | |
High | 290 (20) | 201 (44) | 491 (26) | |
Gleason score, n (%) | ||||
2–6 | 596 (42) | 132 (29) | 728 (39) | <0.001 |
3+4 | 378 (26) | 99 (22) | 477 (25) | |
4+3 | 239 (17) | 91 (20) | 330 (17) | |
8–9 | 197 (14) | 129 (28) | 326 (17) | |
Not reported | 18 (1) | 9 (2) | 27 (1) | |
T stage, n (%) | ||||
T1–2a | 457 (32) | 127 (28) | 584 (31) | 0.17 |
T2b/c/x | 450 (32) | 168 (37) | 618 (33) | |
Not reported | 521 (36) | 165 (36) | 686 (36) | |
PSA level (ng/mL), n (%) | ||||
0–10 | 737 (52) | 131 (28) | 868 (46) | <0.001 |
>10–20 | 383 (27) | 125 (27) | 508 (27) | |
>20 | 132 (9) | 124 (27) | 256 (14) | |
Not reported | 176 (12) | 80 (17) | 256 (14) |
Curative-intent treatments (RP and radiotherapy) were used more frequently in the 70–79 years group (62%) compared with the older group (18%). There was a higher rate of RP in the 70–79 years group compared with the older group (18% vs 0%). Similarly, the rate of radical radiotherapy was higher in the 70–79 years group (44% vs 18%). In contrast, primary hormonal therapy (4% vs 21%) and watchful waiting (16% vs 22%) were used more frequently in the 80–89 years group (Table 2). Covariate adjusted logistic regression indicated that older men were far less likely to be treated with curative treatment (OR 0.12, 95% CI 0.09–0.16; P < 0.001). Propensity score matching was able to match all 460 older men to younger individuals. Conditional logistic regression indicated that the odds of receiving curative treatment for individuals aged 80–89 years was markedly lower than that of men aged 70–79 years (OR 0.13, 95% CI 0.09–0.19; P < 0.001).
Age group | ||
---|---|---|
Treatment approach | 70–79 years, n (%) | 80–89 years, n (%) |
Number of patients | 1 428 | 460 |
Open RP | 103 (7) | – |
RALP | 160 (11) | – |
EBRT | 561 (39) | 81 (18) |
Brachytherapy | 2 (<1) | – |
HDR | 7 (<1) | – |
LDR | 54 (4) | – |
Anti-androgen | 43 (3) | 64 (14) |
LHRH agonist | 18 (1) | 31 (7) |
Active surveillance | 102 (7) | 11 (2) |
Watchful waiting | 229 (16) | 100 (22) |
Other | 149 (10) | 173 (38) |
- EBRT, External beam radiation therapy; HDR, high-dose-rate brachytherapy; LDR, low-dose rate brachytherapy; RALP, robot-assisted laparoscopic RP.
The duration of follow-up was similar across both two groups (median 7.3 years in the 70–79 years group vs 7.0 years in the 80–89 years group). At time of analysis, 263 (57%) deaths had occurred in the 80–89 years group compared to 310 (22%) in the 70–79 years group. OS was better in the 70–79 years group, with 87% 5-year survival and 68% 10-year survival, vs 55% 5-year survival and 26% 10-year survival in the 80–89 years group (Table 3). In the Cox proportional hazard regression, the estimated increased risk of mortality for the older group compared to 70–79 years group was HR 3.7 (95% CI 3.2–4.4; P < 0.001).
Age group | |||
---|---|---|---|
Variable | 70–79 years | 80–89 years | Total |
Number of patients | 1 428 | 460 | 1 888 |
Cause of death, n (%) | |||
(Alive) | 1 118 (78) | 197 (43) | 1 315 (70) |
Not prostate cancer | 213 (15) | 190 (41) | 403 (21) |
Prostate cancer | 97 (7) | 73 (16) | 170 (9) |
5-year OS, % (95% CI) | 87 (85–88) | 55 (51–60) | 79 (77–81) |
Events/n | 310/1 428 | 263/460 | 573/1 888 |
10-year OS, % (95% CI) | 68 (64–72) | 26 (21–33) | 58 (54–61) |
Events/n | 310/1 428 | 263/460 | 573/1 888 |
Follow-up, years, median (95% CI) | 7.3 (7.1–7.5) | 7.0 (6.8–7.7) | 7.3 (7.1–7.5) |
Events/n | 1 118/1 428 | 197/460 | 1 315/1 888 |
The proportion of the deaths related to prostate cancer was similar for the two age cohorts: of the 263 deaths in the 80–89 years group, 73 (28%) were attributable to prostate cancer, as compared to 97 (31%) of the 310 deaths in the 70–79 years group (Table 3).
Univariate competing risk analyses indicated that the risk of death from prostate cancer was higher in patients with higher Gleason score (Gleason score 7 vs ≤6, sHR 1.4 [95% CI 1.0–2.1], P = 0.06; Gleason score 8–9 vs ≤6, sHR 3.1 [95% CI 2.1–4.7], P < 0.001); higher PSA at diagnosis (PSA level >10–20 vs ≤10 ng/mL, sHR 1.9 [95% CI 1.3–2.9], P = 0.001; PSA level >20 vs ≤10 ng/mL, sHR 3.7 [95% CI 2.4–5.6], P < 0.001), higher risk groups (NCCN intermediate vs low, sHR 1.8 [95% CI 1.1–2.9], P = 0.01; high vs low, sHR 3.6 [95% CI 2.3–5.8], P < 0.001), but not significantly different for higher T-stages (T2b/c vs T1–2a, sHR 1.4 [95% CI 1.0–2.2], P = 0.07). Older patients were at higher risk of death both due to prostate cancer (aged 80–89 vs 70–79 years, sHR 2.5 [95% CI 1.8–3.4], P < 0.001) and due to other causes (aged 80–89 vs 70–79 years, sHR 3.4 [95% CI 2.8–4.2], P < 0.001). Figure 2 presents cumulative incidence PCSM and non-prostate cancer mortality by 5-year age groups and NCCN risk stratification (low/intermediate vs high). Competing-risk regression in propensity score matched individuals (two 10-year age groups: 70–79 vs 80–89 years) indicated that the risk of PCSM was reduced in both 10-year age groups in individuals treated with curative intent (aged 70–79 years, HR 0.57 [95% CI 0.34–0.97], P = 0.04; aged 80–89 years, sHR 0.32 [95% CI 0.13–0.80], P = 0.02; Table 4). The complete case sensitivity analysis indicated the same conclusion for the benefit of curative-intent treatments in older men (aged 70–79 years, sHR 0.61 [95% CI 0.30–1.25], P = 0.18; aged 80–89 years, sHR 0.24 [95% CI 0.07–0.76], P = 0.02).

Non-curative | Curative | |||
---|---|---|---|---|
Age, years | % (E[PCSM]/n) | % (E[PCSM]/n) | sHR (95% CI) | P |
70–79 | 9 (46/505) | 6 (18/299) | 0.57 (0.34–0.97) | 0.04 |
80–89 | 19 (27/142) | 7 (5/72) | 0.32 (0.13–0.80) | 0.02 |
- E[PCa], PCSM events; n, number in matched cohort.
Discussion
There are limited data available on older patients diagnosed with localised prostate cancer aged ≥80 years. The changes in demographics and the increasing number of older men globally, necessitate special attention to the needs of this group. The commonly held belief is that elderly men die with, rather than from prostate cancer.
In both groups, about one-third of the deaths were due to prostate cancer. The older group were more likely to have high-risk disease at diagnosis. Previous studies focusing on older men have shown that elderly men present with more aggressive disease 13. A study by Brassell et al. 14 showed that elderly men (aged ≥70 years) present with higher clinical stage, biopsy grade, and pre-diagnosis PSA velocity. In another study comparing men aged >70 years with younger patients; elderly men had a higher proportion of pathological tumour stage 3/4, Gleason score >7, and larger tumour volume; the older men also had worse disease-free survival and higher risk of PCSM 15. A review of the Surveillance, Epidemiology, and End Results (SEER) database, concluded that older patients (aged ≥75 years), were more likely to present with advanced disease and had a higher risk of PCSM compared with patients aged <75 years 16. We have confirmed that the same trend continues amongst older patients (80–89 years group) who present with significantly higher Gleason score and also higher diagnosis PSA when compared with relatively younger patients (the 70–79 years group). We believe our present results, reinforce the available evidence and expand the age limit of the mentioned studies with a special focus on men aged ≥80 years. These findings could be, at least in part, attributed to delays in treatment or decreased rates of screening in the elderly (lead-time bias). However, high Gleason score, a known inherent indicator of aggressiveness of prostate cancer cells, is unlikely to be influenced by this lead-time bias. In elderly patients, changes in the immune system 17 and the microenvironment 18 may play a role in the prostate cancer presentation and outcome.
In our present study, patients in the older group were significantly less likely to receive curative-intent treatment. This is in accordance with available literature showing that subsets of elderly men with prostate cancer are undertreated 9, 19. Current guidelines recommend that in men with prostate cancer, the decision to offer curative-intent treatment should be based on estimates of life-expectancy 10, 20. This is to avoid overtreatment in men whose life may be limited by other factors and is based on the hypothesis that these patients may die from other causes rather than prostate cancer. Estimating life-expectancy on an individual basis is difficult, especially in the elderly. In addition, retrospective data shows that, even following current guidelines, elderly men with a life-expectancy of >10 years may not receive the recommended curative-intent treatment and men aged 70–80 years with localised prostate cancer are undertreated 7. Our present data highlights that in men aged 80–89 years at diagnosis, the odds of receiving curative-intent treatment is significantly lower even compared with men aged 70–79 years (who are also, likely to be undertreated compared with men aged <70 years at diagnosis). We have shown that when treated curatively, the risk of PCSM is reduced in all age groups.
Electing non-curative approaches in elderly patients with high-risk localised prostate cancer may be based on the presence of comorbidities or patients’ preference. Nevertheless, in some patients, clinician bias may also play a role. Clinician bias is known to affect clinical decision-making 21 and may lead to underutilisation of evidence-based treatment options 22; this is more common when treating elderly patients 23. Clinician bias may be more prominent in areas where there is little data to support an evidence-based approach, so decisions are made based on experience or extrapolation, which increases the risk of bias. The management of early prostate cancer in elderly men is an area where the paucity of reliable data can increase the risk of bias. The assumption that elderly patients are less likely to die of prostate cancer (clinicians bias) may lead to a conservative non-curative approach in this group, despite the fact that some would benefit from a curative approach. Our present study seeks to challenge this assumption.
Our present study is limited by its retrospective, non-randomised design. However, given that conducting a randomised prospective study to compare curative treatment vs non-curative approaches in patients with good performance is ethically challenging, retrospective analyses of good quality data remain one of the main sources of information to guide management of this group. In the future, Prostate Cancer Outcomes Registry-Australia and New Zealand (PCOR-ANZ) will further assist in collecting data and will provide an evidence base in this area 24. One limitation of our present study is the lack of data regarding patient performance status. It is possible that patients who received curative-intent treatment in our present cohort had better performance status. Another potential source of bias is inaccurate reporting of cause of death. The SA death registry gathers information on cause of death from death certificates. Previous studies have shown that this can be a reliable but not perfect source of information in determining the cause of death in patients with prostate cancer 25. Verifying the precise cause of death from other sources is not practical for a large population-based registry such as the SA-PCCOC.
We would like to emphasise that the results of our present study should not be interpreted as advocating aggressive curative treatment in all elderly patients, where the risk of over-treatment outweighs the risk of under-treatment. Rather we provide evidence, supporting the hypothesis that there is a group of elderly who may potentially benefit from curative-intent treatments. Accurately identifying this group is challenging. A multidisciplinary approach has been shown to impact on the decision-making process 26, 27 and can help decrease the risk of over-treatment or under-treatment in these patients.
The present study shows that a considerable proportion of men diagnosed with localised prostate cancer aged ≥80 years die from the disease. Our present findings support a carefully individualised approach to consider curative-intent treatments in selected elderly patients with localised prostate cancer: an approach based on biological age rather than chronological age. We suggest that treatment options including curative options should be discussed with patients with high-risk features, who have a reasonable life-expectancy (based on comorbidities) regardless of their age. Chronological age should not be the sole factor leading to the exclusion of curative approaches in the discussion with patients. The decision-making process should obviously follow the patient's wishes and involve his family, and should include a thorough assessment of patient's performance status and comorbidities. The process may require geriatric assessments and a multidisciplinary approach. We propose future research aimed at prospectively identifying elderly men whose prognosis could be improved with curative-intent treatments.
Funding
No specific funding for this work to report: This work was supported by SA Prostate Cancer Clinical Outcomes Collaborative.
Conflict of Interest
The authors declare no conflict of interest.
References
Abbreviations
-
- NCCN
-
- National Comprehensive Cancer Network
-
- OR
-
- odds ratio
-
- OS
-
- overall survival
-
- RP
-
- radical prostatectomy
-
- SA-PCCOC
-
- South Australian Prostate Cancer Clinical Outcomes Collaborative
-
- SA
-
- South Australia
-
- sHR
-
- sub-distribution hazard ratio