Biopsy features associated with prostate cancer progression in active surveillance patients: comparison of three statistical models

Authors


Correspondence: Viacheslav Iremashvili, Department of Urology, University of Miami Miller School of Medicine, PO Box 016960 (M-14), Miami, FL 33101, USA.

e-mail: viremashvili@med.miami.edu

Abstract

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

  • Active surveillance is an established management option for patients with favourable-risk prostate cancer. However, about 25–30% of active surveillance patients demonstrate biopsy progression within the first 3–5 years of follow-up. Although several factors, such as the results of the diagnostic and surveillance biopsies, are known to be associated with the risk of progression, our ability to accurately predict this risk remains limited.
  • Our analysis demonstrated that the overall number of positive cores in the diagnostic and first surveillance biopsies is strongly associated with the risk of progression in active surveillance patients. Furthermore, combined results of diagnostic and first surveillance biopsies provide more information about the probability of progression than they do separately. The most important variable affecting the progression-free survival was the overall number of cores positive for cancer. By 3 years of active surveillance, most of the patients who had four positive cores in the diagnostic and surveillance biopsies progressed, while those who had only one positive core had an excellent prognosis. These findings could be used to improve the accuracy of assessments of the prognosis of patients with low-risk prostate cancer and to help them make informed decisions about their treatment.

Objective

  • To analyse the prognostic importance of information provided by the diagnostic biopsy, the first surveillance biopsy and a combination thereof to identify active surveillance patients with a particularly high risk of progression.

Materials and Methods

  • The present study included 161 active surveillance patients who had at least two surveillance biopsies.
  • The first surveillance biopsy was performed within 1 year of the diagnosis. Further surveillance biopsies usually took place every 1–2 years.
  • Progression on the surveillance biopsy was defined as the presence of Gleason 4/5 cancer, > two positive cores or >20% involvement of any core.
  • Cox proportional hazards regression analysis was used to examine the relationship between biopsy characteristics and progression. Three distinct statistical models were built using characteristics of diagnostic biopsies, surveillance biopsies, and a combination thereof. Harrell's c-index was used to quantify the predictive accuracy of each multivariate Cox model.

Results

  • The median follow-up was 3.6 years; 46 (28.6%) patients progressed.
  • In multivariate analysis the major factor associated with progression was the number of positive cores.
  • The model based on the combined results of diagnostic and first surveillance biopsies was significantly more predictive than the models based on the individual results of each biopsy.
  • Patients with four positive cores in the diagnostic and first surveillance biopsies had estimated 5-year progression rate of 100%.

Conclusion

  • The total number of positive cores in the diagnostic and first surveillance biopsies provides important information about the risk of prostate cancer progression in active surveillance patients.
Abbreviations
HGPIN

high-grade prostatic intra-epithelial neoplasia

IQR

interquartile ratio

Introduction

The widespread use of PSA screening during the last two decades has resulted in a dramatic increase in the incidence of prostate cancer [1]. Many prostate cancers are diagnosed at an early stage and, if left untreated, will remain indolent during the patients' lifetime [2]. Therefore, radical treatment of all prostate cancers could result in substantial over-treatment. In recent years, active surveillance has been advocated as a potential solution to overtreatment [3]. This management strategy has become an established alternative for patients with low-risk prostate cancer [4, 5]. Ideally, active surveillance should delay and possibly avoid the morbidity associated with radical treatments while avoiding disease progression [6].

Success of active surveillance generally depends on two pivotal principles. The first principle concerns the accurate selection of patients with favourable prognosis using a certain set of criteria. Patients with low-volume/low-grade cancers are generally considered to be the optimal candidates for active surveillance [7]. The second principle involves close and safe monitoring for any signs of cancer progression beyond the acceptable limits. Both are primarily based on the information provided by prostate biopsies which could be supplemented by PSA level. Although the ability of transrectal prostate biopsy to identify patients for active surveillance has been studied extensively [8, 9], there are still many questions about the limitations of this procedure. For instance, the value of biopsy characteristics in predicting the progression of disease during the follow-up in patients initially eligible for active surveillance is poorly known. We hypothesized that combining the results of diagnostic and first surveillance biopsies might provide additional prognostic information to partially overcome the limitations of the prostate biopsy technique. To test this hypothesis we compared the value of information provided by diagnostic biopsy, the first surveillance biopsy and a combination thereof in predicting prostate cancer progression in our active surveillance cohort.

Materials and Methods

All patients managed by active surveillance at our institution are entered in a prospectively maintained institutional review board-approved database. Our inclusion criteria for active surveillance are biopsy Gleason grade ≤6, ≤ two positive biopsy cores, ≤20% tumour present in any core, and clinical stage T1–T2a. The outside biopsy slides are usually reviewed by an institutional genitourinary pathologist. Clinical stage is assigned by the attending urologist. The final analysis of this characteristic is based on the 2002 TNM system.

Each patient is followed every 3–4 months with a PSA and rectal exam. The first surveillance biopsy is performed within 1 year of the diagnosis. Further surveillance biopsies take place every 1–2 years. Progression on the surveillance biopsy is defined as the presence of high-grade cancer, > two positive cores or >20% involvement of any core. Although changes on repeat biopsy could result from better sampling of the tumour rather than from actual increase in size or Gleason sum, we use the term ‘progression’ for the sake of consistency.

All patients included in the present study underwent at least two surveillance biopsies. These patients were selected to analyse the prognostic significance of features of the initial two biopsies on the risk of progression on further biopsies. From October 1994 through May 2011, 278 patients with prostate cancer enrolled in our active surveillance programme. Of these patients 17 progressed on the first surveillance biopsy, while 26 and 74 have not yet had their first and second surveillance biopsies, respectively. Exclusion of these patients resulted in a study population of 161 patients, for all of whom there was complete information about the diagnostic and surveillance biopsies. This information included the number of cores taken, number of positive cores, percentage involvement of positive cores by cancer, Gleason sum and presence of high-grade prostatic intra-epithelial neoplasia (HGPIN).

For the regression analysis, time was calculated as a period from the date of diagnostic biopsy (for the model based on its results) or first surveillance biopsy (for the models based on the results of this biopsy and the combined model) to biopsy progression, treatment (in patients still eligible for active surveillance) or last follow-up. Cox proportional hazards regression analysis was used to examine the relationship between biopsy characteristics and progression. Three distinct analyses were performed using characteristics of diagnostic biopsies, surveillance biopsies and a combination thereof. These characteristics included number of cores taken, number of positive cores, presence of HGPIN, and mean percentage of core involvement by cancer. The latter was entered into the analysis as a categorical ordinal variable (≤5%, 6–10%, 11–15% and 16–20%). All variables were tested in univariate analysis and included in the multivariate model if they showed a tendency to be statistically significant (P < 0.15). Harrell's c-index was used to quantify the predictive accuracy of each multivariate Cox model. The value of this variable ranges from 0.5 (representing a random prediction) to 1 (representing perfect discrimination). The cumulative incidence of disease progression in subgroups of patients with different numbers of positive cores was estimated using the Kaplan–Meier method and results were compared with the log-rank test. All tests were two-sided with P ≤ 0.05 considered to indicated statistical significance. Stata® version 11.0 was used for all data analyses.

Results

Characteristics of the 161 patients included in the study are listed in Table 1. Eight patients (5%) had a PSA level ≥10 ng/mL and 53 (33%) had a PSA density ≥0.15 at diagnosis. Most patients (96%) had T1c disease; 39 (24%) patients had at least five years of follow-up and 37 (23%) had at least four surveillance biopsies.

Table 1. Descriptive characteristics of the patients
VariableMean ± SDRangeMedianIQR
  1. PSAD, prostate-specific antigen density.
Age at diagnosis, years61.9 ± 8.242.5–78.962.056.7–68.1
PSA at diagnosis, ng/mL5.1 ± 2.70.6–16.74.93.4–6.0
PSAD at diagnosis0.14 ± 0.080.02–0.560.120.08–0.18
Follow-up, years4.1 ± 2.01.3–12.43.72.7–4.9
No. of surveillance biopsies2.9 ± 1.32–922–3

The median (interquartile range [IQR]) follow-up was 3.6 (2.6–4.6) years. A total of 46 (28.6%) patients had pathological progression such as the presence of Gleason 4/5 cancer or an increase in tumour volume (> two positive cores and/or >20% involvement of any core); 11 (24%) of these patients had only Gleason 4/5 cancer while 12 (26%) had only an increase in tumour volume. Both types of adverse changes were noted in 23 (50%) of the patients. A total of 44/46 (95.7%) patients who progressed by our criteria were also not eligible for active surveillance by the Epstein criteria [10]. Median (IQR) time to progression was 3.0 (2.3–3.8) years. Three (2%) patients preferred not to continue active surveillance even though there was no progression.

The initial and first surveillance biopsies generally consisted of at least 10 cores (Table 2). The diagnostic biopsy uniformly showed low-volume cancer. More than 80% of the patients had only one positive core, while >70% had ≤5% cancer involvement. The first surveillance biopsy did not contain cancer in more than half of the patients. The patients with cancer at the first surveillance biopsy typically had <5% involvement in their respective cores.

Table 2. Characteristics of the diagnosis and first surveillance biopsies
 n (%)
Diagnosis biopsy 
Cores sampled 
<1034 (21.1)
≥10127 (78.9)
Positive cores 
One129 (80.1)
Two32 (19.9)
Mean % core involvement 
1–5113 (70.2)
6–1025 (15.5)
11–1510 (6.2)
16–2013 (8.1)
HGPIN13 (8.1)
First surveillance biopsy 
Cores sampled 
<1022 (13.7)
≥10139 (86.3)
Positive cores 
None90 (55.9)
One44 (27.3)
Two27 (16.8)
Mean % core involvement 
1–537 (23.0)
6–1013 (8.1)
11–1510 (6.2)
16–2011 (6.8)
HGPIN11 (6.8)

Results of the univariate and multivariate Cox regression analyses for the three studied models are shown in Table 3. In all models, both the number of positive cores and the mean percentage core involvement with cancer were significantly associated with increased risk of progression in a univariate analysis. However, only the number of positive cores remained statistically significant in a multivariate analysis, although in the combined model the mean core involvement was of borderline significance. Graphical analysis of the impact of the overall number of positive cores in both biopsies showed only a small difference between patients with two and three positive cores (Fig. 1). Therefore, these two categories were combined in the multivariate model. Harrell's c-index was higher in the multivariate model including the combined results of diagnostic and first surveillance biopsies than it was in the models based on the individual results of each biopsy.

Figure 1.

Progression-free Kaplan–Meier survival curves for sub-groups of patients with different overall numbers of positive cores in the diagnostic and first surveillance biopsies. One core vs other groups, P < 0.001; four cores vs other groups, P < 0.001; two cores vs three cores, P < 0.706.

Table 3. Results of the univariate and multivariate Cox regression analyses for different regression models
VariablesUnivariate analysisMultivariate analysis
HR (95% CI)PHR (95% CI)P
  1. *For the multivariate model. †Groups with two and three positive cores are combined into one category. ‡In any of the two biopsies.
Diagnostic biopsy (Harrel's c-index = 0.694)*
Number of cores taken 0.638  
<101
≥100.852 (0.438–1.657)
Number of positive cores3.487 (1.898–6.409)<0.0013.296 (1.772–6.130)<0.001
Mean % core involvement1.333 (1.037–1.714)0.0251.356 (1.044–1.760)0.022
HGPIN3.096 (1.304–7.350)0.0102.341 (0.905–6.056)0.079
First surveillance biopsy (Harrel's c-index = 0.692)*
Number of cores taken 0.778  
<101
≥101.131 (0.480–2.670)
Number of positive cores2.272 (1.582–3.263)<0.0012.003 (1.113–3.603)0.020
Mean percent core involvement1.544 (1.252–1.905)<0.0011.108 (0.769–1.597)0.619
HGPIN1.485 (0.586–3.760)0.405  
Combination of the two biopsies (Harrel's c-index = 0.751)*
Number of cores taken 0.551  
<201
≥201.230 (0.623–2.428)
Number of positive cores2.329 (1.786–3.035)<0.0012.070 (1.526–2.807)<0.001
Mean % core involvement1.589 (1.234–2.046)0.0011.355 (0.991–1.809)0.064
HGPIN2.016 (0.999–4.070)0.0503.342 (1.587–7.038)0.001

The actual rates of progression-free survival on active surveillance were 85% (69/81) in patients with one positive core, 67% (44/66) in those with two or three positive cores, and 14% (two of 14) in those with four positive cores. Estimated progression-free survival rates followed the same pattern (Table 4).

Table 4. Estimated progression-free survival in subgroups of patients with different total numbers of positive cores in the diagnosis and first surveillance biopsies
Total number of positive coresProgression-free survival, % (95% CI)
At 2 years*At 4 years*
  1. *From the first surveillance biopsy.
194.3 (85.5–97.8)86.1 (73.6–93.0)
2–378.8 (66.3–87.1)59.6 (43.6–72.4)
453.9 (24.8–76.0)0

Discussion

One of the major obstacles to a large-scale implementation of active surveillance is the uncertainty of its long-term efficacy [5]. The available literature indicates a range of progression from one-quarter to one-third of patients over a limited follow-up not exceeding 3–4 years in most published cohorts [11-14]. In the present study, we analysed the prognostic importance of prostate biopsy features to identify a subgroup of active surveillance patients with a particularly high risk of progression.

We performed three sets of multivariate regression analyses and compared the predictive performance of the resulting models. Each model had a specific clinical rationale. The results of the diagnostic biopsy are available when active surveillance is initially considered, and it is an optimal time for clinicians to estimate prognosis and discuss treatment options. On the other hand, the results of the first surveillance biopsy could reflect the dynamic changes of the tumour over the elapsed surveillance time. Therefore, the surveillance biopsy could theoretically be more predictive than the diagnostic biopsy of the possibility of progression. However, the technique of transrectal prostate biopsy, even in its extended version, has limitations [15] and a considerable number of patients who fulfil even the most stringent active surveillance criteria harbour unfavourable prostate cancer characteristics [16]. The combined features of two successive biopsies (which could be looked at as one extensive procedure) could provide more accurate information about the extent of cancer in the prostate. Furthermore, in our practice, most patients are referred from other institutions, so the first surveillance biopsy is often the starting point of our management.

The major outcome of the present study is that the combined results of diagnostic and first surveillance biopsies provide more information about the probability of progression than they do separately. Our analysis has shown that three distinct groups of patients could be defined based on the combined results of the two biopsies. The highest chance of progression was in patients who had two positive cores in each biopsy session. These patients had the worst outcomes with 12/14 progressing over the period of follow-up and an estimated 5-year (4 years of the first surveillance biopsy) progression rate of 100%. Since the prostate biopsy essentially provides only probabilistic information about the actual tumour and tends to underestimate its size and grade, these results are not completely unexpected. If prostate biopsy results consistently show tumour characteristics near the limits of active surveillance eligibility, the inherent risk of progression beyond this boundary can be quite significant.

The best prognosis was demonstrated in patients who had one positive core at diagnosis followed by a negative first surveillance biopsy. Less than 20% of these patients were estimated to progress during 5 years of active surveillance. This confirms earlier findings that one positive core in the diagnostic biopsy and a negative surveillance biopsy are independently associated with good prognosis [14, 17, 18]. However, our analysis has further shown an interaction between these two factors, i.e. that the results of the diagnostic biopsy affect the outcomes in patients with negative first surveillance biopsy. More specifically, patients with two positive cores in the first biopsy showed intermediate risk of progression despite having a negative first surveillance biopsy. This risk group also contains patients with two positive biopsies but fewer than four positive cores overall. The probability of progression in these patients was similar to the mean risk for the entire cohort. This risk group also contains patients with two positive biopsies but fewer than four positive cores overall. The probability of progression in this group was similar to the mean risk for the entire cohort.

Interestingly, the percentage core involvement was not independently associated with the risk of progression in the multivariate analysis. This differs from other studies [17]. A potential explanation for this controversy is the difference in the active surveillance criteria used. Our criteria are conservative with regard to maximal core involvement, and therefore the variability of this variable in the present cohort was significantly limited.

The non-extended prostate biopsy scheme is known to provide less accurate information about the presence or absence of cancer and its characteristics [19]. However, the type of biopsy was not statistically significantly associated with the probability of progression in all of the studied models. This probably resulted from the small number of non-extended biopsies performed in our cohort.

High-grade prostatic intra-epithelial neoplasia is a well-established precursor to prostate cancer, but its clinical importance, e.g. regarding the risk of finding prostate cancer at rebiopsy, is controversial [20]. While HGPIN in individual biopsies was not associated with progression in our analysis, it was shown to be an independent risk factor in the multivariate analysis of the combined model. Since only a few patients in our cohort had HGPIN, this finding needs to be tested in further studies.

The limitations of the present study include its retrospective nature as well as the moderate sample size and duration of the follow-up. In the present study we focused only on the biopsy results and did not analyse other variables potentially associated with the progression such as PSA level and its derivatives. Also we did not use any biomarkers, some of which were recently shown to be associated with biopsy progression in patients managed by active surveillance [21]. Most of the diagnosis biopsies were performed in other institutions due to the referral nature of our practice. This resulted in certain heterogeneity of the biopsy techniques. However, all of these procedures were systematic transrectal biopsies and most included at least 10 cores. Furthermore, we believe that it makes our study cohort more representative of the ‘real world’ urology practice.

The selection of patients for our active surveillance cohort was based on prostate biopsy criteria that are more restrictive than those used by other authors. Also we do not use PSA or its derivatives to establish the eligibility of patients for active surveillance. These differences could affect the generalizability of the results of the present study. As mentioned earlier, these discrepancies could be particularly important for the effect of percentage core involvement on the probability of progression. On the other hand, most of the study patients who demonstrated progression on surveillance biopsies by our criteria would also not be eligible for active surveillance by the widely used Epstein criteria. Furthermore, we believe that the correlation between the number of positive cores in the two biopsies and prostate cancer progression primarily results from the limitations of transrectal biopsy. Therefore, our findings could hold true for other active surveillance cohorts in which this technique is used for patient selection and follow-up.

In conclusion, combined results of the diagnostic and first surveillance biopsies are significantly more predictive of the risk of progression than the features of each individual biopsy. The most important variable affecting the progression-free survival in our cohort was the overall number of cores positive for cancer. Most of the patients with the maximal number of positive cores in the first two biopsies have shown signs of progression on further surveillance biopsies. On the other hand, patients with only one positive core in the diagnostic biopsy and a negative first surveillance biopsy had an excellent prognosis. These findings could be used to assess the prognosis of patients with low-risk prostate cancer and to help them make informed decisions about their treatment.

Acknowledgements

The authors thank Center for Urologic Research, Education, and Diseases (CURED) and Mr Vincent Rodriguez.

Conflict of Interest

None declared. Source of funding: Department of Urology, University of Miami Miller School of Medicine.

Ancillary