SEARCH

SEARCH BY CITATION

Keywords:

  • prostate cancer;
  • non-curative;
  • active surveillance;
  • prostatectomy;
  • radiation;
  • undergrading

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Funding and Acknowledgements
  8. Conflict of Interest
  9. References

What's known on the subject? and What does the study add?

  • The recently published Prostate Cancer Intervention versus Observation Trial (PIVOT) did not identify differences in prostate cancer-specific mortality or all-cause mortality among patients with low-risk disease managed conservatively vs those managed definitively; however, recently published data suggest that older men may harbour more aggressive disease than is identified at biopsy owing to sampling error and undergrading. Whether older men with apparent low-risk disease are placed at risk of prostate cancer-specific mortality when managed conservatively remains unknown.
  • The study used population-level data to show that non-curative approaches for older men with low-risk prostate cancer do result in an increased risk of prostate cancer-specific mortality. Differences between our study and the PIVOT trial include the fact that we included a larger sample size, analysed the data using an ‘as-treated’ approach, and included a healthier cohort of men as evinced by lower 4-year all-cause mortality estimates in our study than in the PIVOT. Our results suggest that older men with apparent low-risk prostate cancer are at risk of undergrading, which probably explains the differences in prostate cancer-specific mortality observed between men managed conservatively vs those managed definitively. Our study suggests that alternative approaches to excluding occult, high grade prostate cancer are needed in such men.

Objective

  • To evaluate whether older age in men with low-risk prostate cancer increases the risk of prostate cancer-specific mortality (PCSM) when non-curative approaches are selected as initial management.

Patients and Methods

  • The study cohort consisted of 27 969 men, with a median age of 67 years, with prostate-specific antigen (PSA)-detected, low-risk prostate cancer (clinical category T1c, Gleason score ≤6, and PSA ≤10) identified by the Surveillance, Epidemiology and End Results programme between 2004 and 2007.
  • Fine and Gray's competing risk regression analysis was used to evaluate whether management with non-curative vs curative therapy was associated with an increased risk of PCSM after adjusting for PSA level, age at diagnosis and year of diagnosis.

Results

  • After a median follow-up of 2.75 years, 1121 men died, 60 (5.4%) from prostate cancer.
  • Both older age (adjusted hazard ratio [AHR] 1.05; 95% confidence interval (CI) 1.02–1.08; P < 0.001) and non-curative treatment (AHR 3.34; 95% CI 1.97–5.67; P < 0.001) were significantly associated with an increased risk of PCSM.
  • Men > the median age experienced increased estimates of PCSM when treated with non-curative as opposed to curative intent (P < 0.001); this finding was not seen in men ≤ the median age (P = 0.17).

Conclusion

  • Pending prospective validation, our study suggests that non-curative approaches for older men with ‘low-risk’ prostate cancer result in an increased risk of PCSM, suggesting the need for alternative approaches to exclude occult, high grade prostate cancer in these men.

Abbreviations
PCSM

prostate cancer-specific mortality

HR

hazard ratio

AHR

adjusted hazard ratio

PIVOT

Prostate Cancer Intervention versus Observation Trial

SEER

Surveillance, Epidemiology and End Results

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Funding and Acknowledgements
  8. Conflict of Interest
  9. References

Screening using an annual prostate-specific antigen (PSA) serum test results in the overdiagnosis and overtreatment of prostate cancer [1]. Specifically, in cancers detected by PSA-based screening, the number needed to treat in order to prevent one death secondary to prostate cancer has ranged from 33 in the European Randomized Study of Screening for Prostate Cancer [2], to 12 in the Swedish trial [3], to five in men with no life-shortening comorbidities in a post-randomization, hypothesis-generating analysis of the Prostate, Lung, Colorectal, and Ovarian cancer screening trial on prostate cancer [4]. As a result, there has been increased use of initially non-curative management approaches such as active surveillance and focal ablation in men with low-risk prostate cancer, particularly in older men, who are often at greater risk of dying from competing risks than from prostate cancer [5, 6]. Furthermore, as a result of these data, the National Institutes of Health recently endorsed active surveillance as an option for all men with low-risk prostate cancer [7].

Although men ≥75 years of age comprise only 26% of newly diagnosed prostate cancer cases, they contribute 53% of deaths secondary to prostate cancer [8]. The underlying aetiology of the increased risk of mortality attributable to prostate cancer in older men has not been definitively elucidated, but it is known that older age is associated with the presence of high grade prostate cancer [9, 10], which may be missed because of the known sampling error associated with prostate biopsy, even when extended (10–14-core) techniques are used [11-14]. This is of particular concern in older men, where BPH is more prevalent, as BPH increases gland volume which in turn increases the number of sampling errors associated with prostate biopsy [10, 15]. Recent data supporting this hypothesis are provided by a study that showed that upgrading at radical prostatectomy was significantly higher in men aged >70 years with very low-risk prostate cancer [16] compared with younger men [17]. It is plausible that older men may be at increased risk of prostate cancer-specific mortality (PCSM) if occult high grade disease is missed and not definitively treated.

We undertook a population-based study using the Surveillance, Epidemiology, and End Results (SEER) database to ascertain whether men of advanced age with PSA-detected and low-risk prostate cancer were at increased risk of PCSM when initially non-curative vs curative management approaches were used.

Patients and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Funding and Acknowledgements
  8. Conflict of Interest
  9. References

Patient Population

We used a population-based registry, SEER [18], in order to identify 27 969 men with non-palpable, PSA-detected, low-risk prostate cancer [16], defined by a clinical classification of T1c (tumour identified by needle biopsy, e.g. because of elevated PSA level) with no tumour identified by DRE), a Gleason score ≤6, and a pretreatment PSA level ≤ 10 ng/mL, diagnosed between 2004 and 2007. The inclusion period was limited to 2004–2007, as PSA data were not available before 2004 and because 2007 represents the most recent full year for which complete information was available on all men. The Gleason score variable, as provided by the SEER programme, indicates the highest Gleason score identified at either biopsy or surgery. Sponsored by the National Cancer Institute, the SEER programme collects and publishes cancer incidence, treatment and survival data from population-based cancer registries; the programme captures ∼97% of incident cancers and the 17 tumour registries cover ∼26% of the US population. Registries report information on age, date of diagnosis, demographics, tumour characteristics, surgical treatment, radiation therapy, overall survival and cancer-specific survival [18]. The number of cores obtained at biopsy is not recorded by SEER.

Patient Management

Initial management approaches were classified as either curative or non-curative. Curative therapies included those considered to represent standard initial treatment options, such as radical prostatectomy, with or without the use of radiation therapy, and radiation therapy (either external beam or brachytherapy), with or without the use of hormonal therapy. Non-curative initial management approaches included coding from SEER that would encompass active surveillance or watchful waiting (n = 8896; this group cannot be further subcategorized using SEER), TURP (n = 201), local tumour excision (n = 31), subtotal or segmental prostatectomy (n = 4), hyperthermia as monotherapy (n = 4), laser ablation (n = 44) and cryosurgery (n = 22). Cryosurgery was deemed non-curative because it is not considered to be a standard initial therapy option by the National Comprehensive Cancer Network [19]. The present study was approved by the institutional review board at our institution. Written consent was not obtained from participants; a waiver was employed.

Follow-up and Determination of Cause of Death

The primary outcome in the present study was the risk of PCSM. The cause of death was determined from the SEER database. Previous studies [20, 21] have demonstrated very high agreement (87–97%) between the cause of death listed in the SEER database and the cause of death determined through medical record review by experts. The cut-off date for determination of death was 31 December 2008. The minimum and median follow-up for patients in the study were 1 month and 2.75 years, respectively.

Statistical Methods

Table 1 shows the distribution of the clinical characteristics at diagnosis of the 27 969 men in the study cohort, as well as the characteristics of men aged > vs those aged ≤ the median age. Comparisons between age groups were performed using the chi-squared test for categorical variables and the Wilcoxon rank-sum test for continuous variables.

Table 1. Clinical factors for men in the study cohort, as stratified by median age at diagnosis
Clinical characteristicWhole cohort: N > 27 969Age ≤ 67 years: N > 14 772Age > 67 years: N > 13 197P*
  1. IQR, interquartile range. *P values compare both age groups. Percentages may not add up to 100 due to rounding.

Median (IQR) age at diagnosis, years67 (61–73)61 (57–67)73 (70–76)
Year of diagnosis, n (%)*   0.002
20046 387 (22.8)3 301 (22.3)3086 (23.4) 
20056 201 (22.2)3 203 (21.7)2998 (22.7) 
20067 598 (27.2)4 029 (27.3)3569 (27.0) 
20077 783 (27.8)4 239 (28.7)3544 (26.9) 
Median PSA (IQR), ng/mL5.6 (4.6–7.1)5.4 (4.4–6.7)6.0 (4.8–7.5)<0.001
Management approach , n (%)*   <0.001
Curative18 767 (67.1)10 298 (69.7)8469 (64.2) 
Non-curative9 202 (32.9)4 474 (30.3)4728 (35.8) 

We used Fine and Gray's univariable and multivariable competing risks regression to assess whether age at diagnosis, year of diagnosis, PSA level and initial management approach (i.e. curative vs non-curative) were associated with the risk of PCSM, the primary outcome measure for the present study. Overall survival was not used as a primary outcome measure because data relating to comorbidity are not available in SEER. PSA level was log-transformed to ensure that this variable followed a normal distribution. Patients with missing data were excluded from the analysis. The analysis was repeated after excluding men who underwent radical prostatectomy as all or part of their initial treatment course, and was also repeated after excluding men who underwent minimally invasive treatment (so that the non-curative arm only contained patients managed with active surveillance or watchful waiting). Unadjusted hazard ratios (HRs) and adjusted HRs (AHRs) with their associated 95% CIs were calculated for each covariate. A two-sided P value of <0.05 was considered to indicate statistical significance.

Cumulative incidences of PCSM were calculated, stratified by clinical factors that were significantly associated with the risk of PCSM from the competing risks regression model, and graphically displayed. Specifically, estimates of PCSM were stratified by: (i) Initially curative vs non-curative management approach; (ii) age > median age vs ≤ median age; and (iii) both the initial management approach and age relative to the median. Point estimates of PCSM and associated 95% CIs were calculated. Comparisons of these estimates were made using Gray's k-mean P value. The Bonferroni correction was used to adjust for multiple (n > 4) comparisons. R version 2.12.0 software (R Foundation for Statistical Computing, Vienna, Austria) was used for all calculations pertaining to Gray's k-mean P value and Fine and Gray's regression. SAS version 9.3 software (SAS Institute, Cary, NC, USA) was used for all remaining analyses.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Funding and Acknowledgements
  8. Conflict of Interest
  9. References

Clinical Characteristics of the Study Cohort

As shown in Table 1, the median (interquartile range) age and PSA level of the 27 969 men who comprised the study cohort were 67 (61–73) years and 5.6 (4.6–7.1) ng/mL, respectively. A total of 18 767 (67.1%) of the 27 969 men were managed with curative intent. The median PSA level was significantly higher in men > the median age (median 5.4 and 6.0 ng/mL for patients ≤67 and >67 years of age, respectively, P < 0.001). The proportion of men initially managed with curative treatment was lower for men > the median age (69.7% vs 64.2% for patients ≤67 and >67 years of age, respectively, P < 0.001).

Factors Associated with the Risk of PCSM

After a median follow-up of 2.75 years, 1121 men had died, of whom 60 (5.4%) had died from prostate cancer. The 4-year Kaplan– Meier estimates of all-cause mortality in patients managed with curative and non-curative initial therapy were 4.33% (95% CI 3.95–4.76%) and 9.27% (95% CI 8.48%–10.14%), respectively. Both older age (AHR 1.05; 95% CI 1.02–1.08; P < 0.001) and the use of non-curative treatment initially (AHR 3.34; 95% CI 1.97–5.67; P < 0.001) were significantly associated with an increased risk of PCSM (Table 2). When excluding men treated with radical prostatectomy, the results were similar, as use of non-curative treatment as initial therapy was associated with increased risk of PCSM (AHR 3.25; 95% CI 1.92–5.05; P < 0.001). We also repeated the analysis after excluding men who received minimally invasive treatments, so that the non-curative cohort contained only patients managed with active surveillance or watchful waiting, and the association between the use of non-curative therapy and PCSM did not meaningfully change (AHR for non-curative vs curative therapy 3.40; 95% CI 1.97–5.85; P < 0.001).

Table 2. Competing risks regression analysis for the risk of prostate cancer-specific mortality for all men
CharacteristicNo. of menNo. of eventsUnivariable analysisMultivariable analysis
HR95% CIPAHR95% CIP
Whole cohort27 96960      
Age at diagnosis, per year increase  1.071.03–1.10<0.0011.051.02–1.08<0.001
Year of diagnosis, per year from 2004  0.890.67–1.170.400.870.66–1.150.33
Log-PSA, per log-ng/mL  0.910.50–1.660.760.840.51–1.380.49
Management approach        
Curative  1.0 (ref)  1.0 (ref)  
Non-curative  3.712.20–6.26<0.0013.341.97–5.67<0.001

Cumulative Incidences of PCSM

Cumulative incidence estimates of PCSM stratified by curative vs non-curative initial management approach, age > median age vs ≤ median age, and the combination of initial management approach and age as stratified around the median are shown in Fig. 1A–C, respectively. Estimates of PCSM were significantly higher, within the entire study cohort, in men who received initially non-curative as compared with curative approaches (P < 0.001, Fig. 1A), as well as in men whose age was > the median age (67 years) as compared with men whose age was ≤ the median age (P < 0.001, Fig. 1B). Men whose age was > the median age also had higher estimates of PCSM when initially non-curative management approaches were used (P < 0.001); this finding was not seen in men whose age was ≤ the median (P = 0.17, Fig. 1C).

figure

Figure 1. Cumulative incidence of PCSM and associated Gray's k-mean P value for all men stratified by initial management with curative or non-curative intent (P < 0.001). A, age > median age vs ≤ median age (P < 0.001) B, and the combination of initial management approach and age as stratified around the median (P < 0.001 for initial non-curative vs curative approach in men >67 years and P = 0.17 for initial non-curative vs curative approach in men ≤67 years). C, Rx, management approach.

Download figure to PowerPoint

The 3-year estimates of PCSM in those managed with curative vs non-curative approaches, respectively, were 0.15% (95% CI 0.09–0.23%) and 0.49% (95% CI 0.34–0.68%), P < 0.001. Respective estimates for men aged ≤67 and >67 years were 0.15% (95% CI 0.08–0.25%) vs 0.38% (95% CI 0.28–0.52%), P < 0.001. Among men aged ≤67 years, 3-year estimates of PCSM in those managed with curative vs non-curative approaches, respectively, were 0.13% (95% CI 0.06–0.25%) vs 0.18% (95% CI 0.07–0.39%), P = 0.17; the respective estimates for men aged > 67 years were 0.17% (95% CI 0.09–0.31%) vs 0.76% (95% CI 0.52–1.09%), P < 0.001.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Funding and Acknowledgements
  8. Conflict of Interest
  9. References

In the present study, using a population-based registry, the initial use of non-curative management and older age were associated with a 3.34-fold and a 5%/year increase in the risk of PCSM, respectively. When the cohort was analysed by age-based subgroup, only men > the median age of 67 years had a significant increase in the cumulative incidence of PCSM when initially treated with a non-curative as compared with a curative approach, despite a median follow-up of only 2.75 years. The clinical significance of these findings is that older men with ‘low-risk’ prostate cancer may harbour occult high grade prostate cancer that can be missed because of prostate biopsy sampling error despite extended (10–14-core) biopsy, placing them at increased risk of PCSM within a few years of the diagnosis when non-curative treatments such as active surveillance are used initially. This is of particular concern because it was standard practice during the study period, and also today, to perform an extended-core biopsy before recommending active surveillance, as well as in surveillance biopsy schemes for patients managed with active surveillance. Data from the Cancer of the Prostate Strategic Urologic Research Endeavor (CaPSURE) database indicate that, in the year 2004, the mean number of cores obtained at biopsy in men with newly diagnosed prostate cancer was 10.2, a number which had risen with time; in 1995, the mean number of cores obtained was only 6.9 [22].

In contrast to older men, younger men are at markedly lower risk of undergrading [15]. In the present study, the nonsignificant association between non-curative management approaches and the cumulative incidence of PCSM in men aged ≤67 years (P = 0.17) suggests that less undergrading is occurring in younger than in older men; however, the present analysis is significantly limited by a small number of events (n = 16) in this cohort of men ≤67 years of age. To better assess the impact of curative therapy on this cohort, a larger sample size and longer follow-up are required.

Several points warrant additional discussion. First, a possible explanation for the association between increasing age and the risk of PCSM may be derived from the established relationship between older age and high grade disease, as seen in the Prostate Cancer Prevention Trial [9]. Undergrading of prostate cancer owing to limitations in current biopsy techniques [12-14] may, therefore, be particularly consequential in older men, in whom undergrading is known to be more common owing to increasing gland volume coupled with increasing prevalence of BPH [15]. Moreover, even among the most favourable setting of ‘very low-risk’ prostate cancer [19], (defined as a clinical classification of T1c, Gleason score ≤6, PSA < 10 ng/mL, PSA density < 0.15 ng/mL/g, <3 positive cores and ≤50% of each core containing prostate cancer) level two evidence indicates that the risk of unsampled Gleason ≥ 7 disease is 26% [23]. More recent data indicate that patients aged ≥ 70 years with very low-risk prostate cancer have a 38% chance of upgrading to Gleason 7–10 disease at the time of prostatectomy, while those who are <70 years have only a 17–19% chance of such upgrading [17]. Thorough review of level one [11, 24-26] and level two [12-14, 27, 28] evidence indicates that extended (10–14-core) prostate biopsy, when compared with sextant biopsy, improves the sensitivity for detection of both prostate cancer and of Gleason score ≥ 7 disease [12-14]; therefore, although the optimum number of cores performed during prostate biopsy remains to be determined, the results of this study suggest that obtaining a median of 10.2 cores for men at initial diagnosis and extended biopsy for those electing active surveillance is inadequate in men aged > 67 years to definitively exclude occult high grade disease in all of these men. It is important to note that obtaining a greater number of cores at biopsy may be associated with increased bleeding [29], although infection rates do not appear to be significantly different [30].

Prospective series of patients managed with active surveillance have shown very low rates of PCSM at 10 years [5, 31], and data from the recently published Prostate Cancer Intervention versus Observation Trial (PIVOT) did not identify differences in PCSM between low-risk patients managed conservatively vs definitively [32]. The present study, which included 27 969 men, had greater power to detect a difference in PCSM than did PIVOT. Additionally, in the PIVOT trial, ∼20% of patients in the observation arm received definitive therapy, and the analysis was performed using intention-to-treat. Although analysis by intention-to-treat is most appropriate for a randomized trial given that randomization is preserved, such an approach further compromises the power of the study to detect a difference in PCSM. The use of an ‘as-treated’ analysis in the present study increased our power to detect such a difference. Moreover, 4-year all-cause mortality estimates in patients managed with curative and non-curative intent in the present study (4.33 and 9.27%, respectively) were lower than respective estimates in the PIVOT trial (9.6 and 14.2%, respectively). We are encouraged by the fact that our overall mortality rates at an early timepoint (i.e. 4 years) when compared with the PIVOT data are numerically lower which may have allowed us, despite shorter follow-up, to be able to see a difference in PCSM across treatment arms (whereas PIVOT did not), given the probable healthier nature of our patient population. Unlike the series published by Tosoian et al. [5] and Klotz et al. [31], not all patients in the present study were managed with active surveillance, and some (who received watchful waiting) may not have been screened for progression of disease and treated definitively when such progression occurred. The inclusion of patients managed with watchful waiting potentially accounts for the ability of the present study to identify a difference between curative and non-curative management in patients with low-risk prostate cancer.

Several potential limitations of the present study warrant mention. First, data relating to comorbidity were not available in the database; however, data from the National Vital Statistics System indicates that the lifespan of an average 70-year-old American male is 14.2 years [33]. Based on the present study results, which show an increased risk of PCSM after a median follow-up of only 2.75 years in men > 67 years initially managed with non-curative approaches, it is reasonable to conclude that such men with no or minimal comorbidity and apparent PSA-detected, ‘low-risk’ disease should be considered for more extensive sampling approaches before recommending non-curative management approaches. In addition, the results of the present study suggest that the mortality associated with undergrading of prostate cancer in older men remains significant despite the increased comorbidity associated with aging. Second, because of the retrospective nature of the present study, it is possible that selection bias could account in part for the association of older age and the increased risk of PCSM that we observed. Specifically, if older men had more advanced disease not captured by the low-risk categorization (e.g. a high PSA velocity, a large percentage of positive cores, a large number of cores with >50% involvement, or tertiary Gleason score 5 disease) as well as comorbid illness and were therefore not definitively treated, it is possible they could die from prostate cancer at a higher rate than a younger, healthier patient with low-risk prostate cancer who did not have these uncaptured adverse prognostic factors. Such bias can only be controlled for in the setting of a randomized clinical trial with pre-randomization stratification by age and comorbidity. Such a trial would ideally balance both known and unknown confounders, and if powered adequately, could define whether the hypothesis we are testing of older age and undersampling of occult high grade disease is true. Third, the median follow-up of the present study was 2.75 years. Yet, despite the short follow-up, significant differences in PCSM in men treated with curative and non-curative approaches were observed, although such differences were small in magnitude; longer follow-up is required to determine whether the magnitude of the differences observed will increase, and also to determine whether younger men with few comorbidities are found to have a significant PCSM reduction with definitive therapy. Fourth, despite the 87–97% concordance [20, 21] between the cause of death identified by the SEER database and by independent medical review, there is the possibility of error in assignment of cause of death in the present study for some men; however we would expect such error to be non-systematic in nature and therefore unlikely to change the associations observed. Finally, given the methodology used by the SEER programme to report Gleason score, in which the highest Gleason score identified at either biopsy or surgery is recorded, patients who were found to have Gleason 6 disease at biopsy but Gleason ≥ 7 disease at prostatectomy were systematically excluded from the analysis, which may make the differences in PCSM between the curative and non-curative arms of the study more pronounced; however, after excluding men who underwent radical prostatectomy, the study findings remained unchanged.

Despite the potential limitations, pending prospective validation, the present study suggests that non-curative approaches for older men with ‘low-risk’ prostate cancer result in an increased risk of PCSM, suggesting the need for alternative approaches to exclude occult, high grade prostate cancer in these men.

Funding and Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Funding and Acknowledgements
  8. Conflict of Interest
  9. References

We would like to thank and acknowledge Dr. Leon Sun for providing the SEER database and for delineating appropriate fields and definitions.

There was no external financial support for this study.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Funding and Acknowledgements
  8. Conflict of Interest
  9. References