We thank Mrs. Helén Ahlgren for assistance in managing the database.
Strategies of active surveillance (AS) of low-risk screen-detected prostate cancer have emerged, because the balance between survival outcomes and quality of life issues when radically treating these malignancies is disputable. Delay before radical treatment caused by active surveillance may be associated with an impaired chance of curability.
Men diagnosed with low-risk (T1c/T2; prostate-specific antigen [PSA] = <10.0; PSA density, <0.2 ng/mL; Gleason score, 3 + 3=6; 1-2 positive biopsies) prostate cancer in the Swedish section of the European Randomized Study of Screening for Prostate Cancer who received radical prostatectomy (RP) were studied. One group received immediate RP, whereas another group received delayed RP after an initial period of expectant management. These groups were compared regarding histopathological and biochemical outcomes, correcting for baseline differences.
Mean follow-up after diagnosis was 5.7 years (standard deviation [SD], 3.2). The immediate RP group (n = 158) received RP a mean of 0.5 (SD, 0.2) years after diagnosis; the delayed RP group (n = 69) received RP after 2.6 (SD, 2.0) years (P < .001). After adjustment for small baseline dissimilarities, no differences in RP frequencies of Gleason score >6 (odds ratio [OR], 1.54; P = .221), capsular penetration (OR, 2.45; P = .091), positive margins (OR, 1.34; P = .445), RP tumor volume (difference, 0.099; P = .155), or biochemical progression rates (P = .185, P = .689) were found between groups, although all data were in favor of immediate RP.
The incidence of small, localized, well-differentiated prostate cancer has risen during the last 2 decades, mainly because of a more widespread use of prostate-specific antigen (PSA) screening.1 Many of these tumors will remain nonharmful during the patient's lifetime, and radical treatment of all these men will result in tremendous overtreatment.2, 3 The significant but modest favorable effect on prostate cancer-specific survival of radical treatment when compared with watchful waiting has to be weighed against the risk of important side effects on an individual basis.4, 5 This issue has become even more important since a large screening trial recently reported a positive effect of screening.6
Strategies of initial active surveillance have emerged that consist of selecting men with a favorable prognosis based on tumor characteristics and initially withholding potentially curative radical treatment, but instead closely monitoring the disease.7 When signs of progression occur, radical treatment is recommended. In retrospective analyses, prostate cancer-specific mortality is very low, and some men die of other causes during follow-up.3 Active surveillance may thus decrease overtreatment of prostate cancer and the risk of side effects by sparing surgery or radiation therapy in some men. In others, however, radical treatment is merely delayed. A crucial question surrounding active surveillance strategies is whether delaying curative treatment is associated with an impaired chance of curability.
In this retrospective study, we compared histopathological and biochemical outcomes of men with screen-detected low-risk prostate cancer between those receiving immediate radical treatment after diagnosis and those receiving delayed radical treatment after an initial period of expectant management.
MATERIALS AND METHODS
Men included in this study all participated in the screening arm of the Swedish section (Gothenburg) of the European Randomized Study of Screening for Prostate Cancer (ERSPC),8 for which they provided written informed consent. The Swedish ERSPC study protocol randomized 20,000 men between 50 and 66 years of age. After approval of the ethical committee in 1994, men randomized to the screening arm have since the beginning in 1995 been offered PSA measurements every 2 years.9 All men with a PSA ≥3.0 ng/mL were candidates for a digital rectal examination, transrectal ultrasound (TRUS), and lateralized sextant prostate biopsies; additional biopsy core(s) were performed in case hypoechoic lesion(s) were seen during TRUS. Prostatic volume was measured by planimetric calculation from the TRUS-recorded measurement of the prostate using the ellipsoid formula. After a prostate cancer diagnosis, decisions on treatment were made after discussing potential treatment options between the physician and the patient, including expectant management if applicable. Treatments were mainly performed in the ERSPC study center itself (Sahlgrenska University Hospital, Gothenburg, Sweden). More detailed information on the ERSPC and the study protocol have been previously published.10
We selected all clinical stage Ic/II (TNM: T1c/T2, N0/X, M0/X) prostate cancer, with a PSA at diagnosis ≤10.0 ng/mL, a PSA density (PSA divided by prostatic volume) <0.2 ng/mL, a Gleason score (pathological dedifferentiation) of 3 + 3 = 6 or more favorable, and 1 or 2 positive biopsy cores. Men with known positive lymph nodes or distant metastases at the time of diagnosis were excluded. The decision to perform a lymph node dissection was made on a patient-specific basis, and consisted of removing all lymphatic tissue in the angle between the obturator nerve and the external iliac vein. This combination of parameters is used as the criteria for eligibility in the international prospective Prostate Cancer Research International: Active Surveillance (PRIAS) study of active surveillance originating from the ERSPC,11 and is largely similar to the inclusion criteria used in the first protocol-based prospective study of active surveillance in Toronto, Canada.12 All patients with prostate cancer with a Gleason sum score <6 were categorized as Gleason 6.
From this group, we selected all individuals who received radical prostatectomy (RP). These patients consisted of 1 group of men who received RP as their initial treatment (immediate RP group) and another group in whom initially an expectant management was elected but who changed to RP later during follow-up (delayed RP group). These 2 groups were compared in this study.
No standardized protocol for expectant management was applied in the delayed RP group, but surveillance was generally based on regular (typically 6 months) PSA measurements, with repeat biopsies in men with increasing PSA, especially in those who preferred to remain on surveillance. Afterwards, RP men were checked every 6 months for PSA. The criteria by Freedland et al were used to define PSA relapse, that is, a PSA value >0.2.13 All prostate biopsy cores and RP specimens were reviewed by the same uropathologist (C.G.P.). Follow-up data were collected from patient charts.
First, we compared baseline characteristics and histopathological outcomes after RP between the immediate RP group and the delayed RP group using the t test for continuous variables and the chi-square test for categorized variables. The following histopathological RP outcomes were assessed: Gleason score >6 (yes/no), capsular penetration (yes/no), positive margins (yes/no), and tumor volume (continuous). As the 2 study groups were not randomized and thus were expected to differ in baseline characteristics, separate logistic and linear regression models were also used for analysis of potential differences in outcome variables. Two separate multivariate analyses were done, 1 to adjust for potential differences at the moment of diagnosis and 1 for the moment of RP. We also assessed whether the time between diagnosis and RP was associated with any of the outcome variables and whether in the delayed RP group the PSA doubling time (PSADT) was associated with any outcomes, using univariate analyses in the 2 separate study groups. Third, we compared time to biochemical progression after RP between groups using Kaplan-Meier, log-rank, and Breslow analysis (tests equality of survival functions by weighting all time points by the number of cases at risk at each time point), using both moment of diagnosis and moment of RP as t = 0, because the immediate RP group likely has a longer follow-up after RP to show biochemical progression, thus possibly introducing a bias. Then, we assessed whether study group was predictive for time to biochemical progression using Cox regression analysis, correcting for differences in variables at the moment of diagnosis as well as at the moment of RP. Finally, we analyzed whether time between diagnosis and RP was predictive for biochemical progression after RP in separate analyses for the immediate RP group and the delayed RP group.
Parameters not further analyzed because of small number of events were mortality (3 men died in the delayed RP group, none of them because of prostate cancer) and seminal vesicle invasion (seen in 1 man in the immediate RP group).
Combining the data of different ERSPC study centers was considered, but rejected because of the small differences in screening protocols that would result in heterogeneity in the study population.
P values (2-sided) <.05 were considered statistically significant. For statistical analysis, the commercially available software Statistical Package for the Social Sciences, version 15.0 (SPSS, Inc, Chicago, Ill) was used.
Table 1 presents study group patient characteristics at the moment of diagnosis and at the moment of RP, follow-up between diagnosis and RP, and histopathological outcomes after RP. Our study group consisted of 227 men who had been diagnosed with low-risk prostate cancer for which they received RP, with a mean follow-up time since diagnosis of 5.7 years. Of these 227 men, 158 (69.6%) primarily elected RP as the initial treatment option, were operated a mean of 0.5 years after diagnosis, and were followed a mean of 5.7 years after diagnosis; 69 (30.4%) first had elected expectant management as the initial treatment option, switched to RP later, were operated a mean of 2.6 years after diagnosis, and were followed a mean of 5.8 years after diagnosis. Of all men diagnosed in the Swedish section of the ERSPC with prostate cancer fulfilling the PRIAS criteria up until the moment of this study analysis, the initial treatments were surveillance in 53%, RP in 42%, and radiation therapy in 5%. An overlap in the distribution of time intervals between diagnosis and RP was seen between men in the immediate RP group (range, 0.1-1.1 years) and men in the delayed RP group (range, 0.6-8.9 years). However, as the intent of treatment was different (ie, first monitor the disease during follow-up before RP vs immediate RP), we maintained this division in initial treatment choice as derived from the patient charts in our analyses. Some men who elected RP as the initial treatment still had a considerable delay (1 patient >1 year) until actual surgery because of patients' delay in making the treatment decision. When compared with patients in the immediate RP group, patients in the delayed RP group had a significantly higher prostate volume and resulting lower PSA density, less positive biopsies, less prostate cancer tissue, higher frequency of nonpalpable (T1c) tumors, higher age and PSA at the moment of RP, longer time between diagnosis and RP, and shorter total follow-up time since surgery. Besides PSA values, specific reasons for switching to RP during expectant management were unavailable for analyses; 65.2% of men in the delayed RP group had a relatively quickly rising PSA, with a PSADT of 0 to 10 years, whereas 34.8% were operated while having a favorably slowly rising or falling PSA. Of all patients suitable for PRIAS who (eventually) received RP, 69% received a lymph node dissection, of which 2% were positive.
Table 1. Immediate RP and Delayed RP Study Groups Baseline Characteristics and Comparisons (N=227)
Total follow-up time since diagnosis, y (mean, median, SD)
5.7, 5.5, 3.2
5.8, 5.4, 3.2
Total follow-up time since surgery, y (mean, median, SD)
5.2, 4.9, 3.1
3.2, 2.1, 3.0
No significant differences were found in the frequency of RP outcome variables Gleason score >6, capsular penetration, positive margins, or tumor size. No RP Gleason sum scores higher than 7 were observed.
Table 2 presents the odds ratios (ORs) for RP Gleason score >6, capsular penetration, and positive margins, and the differences in RP tumor volume, as well as confidence intervals and P values, for the immediate RP versus the delayed RP group. Unadjusted univariate models, adjusted models incorporating the parameters at the moment of diagnosis, and adjusted models also incorporating the parameters at the moment of RP are presented. Adjusted ORs for both multivariate models for Gleason score >6 were 1.54 and 1.47; for capsular penetration, the ORs were 2.45 and 2.48; and for positive margins, the ORs were 1.34 and 1.21; difference in RP tumor volume was 0.099 and 0.008 mL. No significant association of study group (immediate RP vs delayed RP) was found with any of the outcome variables in the unadjusted univariate or in any of the 2 adjusted multivariate models.
Table 2. ORs and Difference in Outcomes of the Delayed RP Group Versus the Immediate RP Group: Unadjusted, Adjusted for Parameters That Are Different Between Groups at the Moment of Diagnosis, and Also Adjusted for Parameters That Are Different Between Groups at the Moment of RPa
Logistic regression was used for categorized variables, and linear regression was used for continuous variables.
Adjusted for tumor volume, millimeters of prostate cancer, and T classification.
Adjusted for volume, millimeters of prostate cancer, T classification, age at RP, and prostate-specific antigen level at RP.
Gleason score >6
Tumor volume, mL
−0.26 to 0.15
−0.058 to 0.36
−0.210 to 0.236
Neither in the immediate RP group nor in the delayed RP group did the time between diagnosis and RP show a univariate significant association with any of the outcome variables. The PSADT between diagnosis and RP not was associated with the any of the outcome variables, neither when stratified in groups (0-3 years, 3-10 years, >10 years, or negative), nor as a continuous variable (with negative values set at 50 years or at 100 years). These findings did not change when only men with ≥3 pretreatment PSA values available were analyzed. When comparing men in the delayed RP group with a short PSADT (0-10 years) at the moment of RP with those with a favorable PSADT (>10 years or negative), no differences in outcomes were found.
Figure 1 presents the Kaplan-Meier curves for biochemical progression after RP in the immediate RP group versus the delayed RP group with the moment of diagnosis as t = 0 (Fig. 1A) and the moment of RP as t = 0 (Fig. 1B). In neither was a significant difference seen between the biochemical progression curves (10-year biochemical progression 9% vs 22% and 9% vs 35%; log-rank P = .185 and .689; Breslow P = .630 and .573, respectively). In Cox regression analysis, study group (immediate RP vs delayed RP) was not a significant predictor of biochemical progression, neither when entering parameters at diagnosis in the model (P = .138, correcting for volume, millimeters of prostate cancer tissue, and T classification) nor when adding parameters to the model at the moment of RP (P = .087, correcting for age at RP, PSA at RP, volume, millimeters of prostate cancer tissue, and T stage). When assessing the immediate RP group and the delayed RP group in separate Cox regression analyses, time between diagnosis and RP was not significantly predictive of biochemical progression after RP in either model.
The above results did not change when we defined our immediate RP and delayed RP groups on the basis of a shorter or longer than 0.5-year delay between diagnosis and RP, instead of the treatment choice as retrieved from the patient charts.
In the limited setting of this study, no significant differences were found in adverse intermediate outcomes after RP between men who received immediate RP and those who received RP after an initial period of expectant management, for small, well-differentiated, localized prostate cancer. Time between diagnosis and RP was not correlated with any of the outcome parameters.
However, with larger patient numbers and longer follow-up, the results of our analysis might have shown significant results. All data seem in favor of the immediate RP group. The most striking number is the almost 2.5× higher odds for capsular penetration after RP in the delayed RP group (Table 2). Still, it should be noted that the ideal study design would include comparisons between the outcomes of the immediate RP group and the outcomes of the delayed RP group plus men who start on expectant management and have stayed on expectant management at the time of analysis. Patients in this virtual third study group did not, however, receive RP and can therefore not be included in our analysis. Whereas in the immediate RP group all men received direct surgery, men in the delayed RP group may have switched to RP because of an unfavorable follow-up, which may be associated with unfavorable outcomes. Specific reasons for the switch are unknown, but 65.2% had a quickly rising PSA (other reasons probably include patient's and physician's desire, anxiety, changes in clinical stage, and/or repeat prostate biopsies). It could then be hypothesized that those men with favorable follow-up, who remain out of consideration in the current study, also have favorable outcomes, and that if all received RP and the results were added to the delayed RP group, this would have a diluting effect on the frequency of unfavorable findings in the delayed RP group. A study design in which the immediate RP and the delayed RP groups are randomized and in which the delayed RP group comprises all men who started on expectant management and received RP after a fixed time interval is difficult to realize.
To illustrate the proportion of men diagnosed with low-risk prostate cancer who switch to active therapy (RP, radiation therapy, hormones) during follow-up after initially starting on an expectant management, Figure 2 presents the treatment-free survival Kaplan-Meier curve of 200 Swedish men diagnosed with low-risk (similar criteria as used in this study) prostate cancer in the ERSPC, who all initially elected expectant management for their disease. Although this is a different group of men than the group that is the main focus of the current study, the proportion of these 200 men who switched to RP during follow-up comprise the current delayed RP group of 69 men. The 10-year treatment-free survival was 40.4%; after 2.6 years (which is the mean follow-up of the delayed RP group in the current study), this was 67.9%. This means that for every man who switched to active therapy, 2.1 men remained on expectant management (32.1% vs 67.9%).
Besides biological tumor progression during expectant management, the potential selection bias as described above may also partly cause differences in unfavorable outcomes between the 2 groups. Our study does not allow for distinguishing in causality between these 2 hypotheses.
In the delayed RP group, side effects of therapy were avoided over a mean period of 2.6 years. In this period, the psychological burden of the disease may have been elevated. The potential favorable effect of expectant management on quality of life because of the avoidance of side effects such as impotence or incontinence must be weighed against the potential difference in outcomes and burden of living with untreated cancer. Anxiety and distress levels of men participating in a prospective active surveillance program have been reported to be favorable.14 However, besides possible worse pathological and biochemical outcomes, delaying radical treatment may also decrease the chance of quality of life-sparing therapies, such as nerve-sparing RP.
The lack of an association between PSA kinetics (PSADT) before RP and outcome variables makes the use of this parameter during active surveillance doubtful. This finding may be counterintuitive, but is in line with the conclusions of a review by Vickers et al, who did not find a predictive value of PSA kinetics beyond PSA alone in untreated patients.15 Fall et al also found that, although being prognostic factors, PSA value and rate of PSA change are both poor predictors of lethal prostate cancer among untreated patients with localized prostate cancer.16 Still, PSADT may prove to hold value in monitoring disease status in men on active surveillance.17
Because of the findings presented above and the knowledge of the mean lead time of >10 years in these low-risk prostate tumors,18 we believe that the harmful effect of initially delaying radical treatment as may be caused by active surveillance protocols is small, which is in line with the conclusion of other reports.19, 20 Furthermore, because the length of delay between diagnosis and RP is not a predictor of adverse outcomes, the treatment of low-risk screen-detected prostate cancer is unlikely to be a matter of emergency. The effect of delay on hard endpoints is, however, for now unknown, although the favorable effect of radical treatment on hard endpoints such as 10-year prostate cancer-specific survival is only 5% in clinically detected prostate cancer.4 Importantly, patients and physicians who choose active surveillance should be aware that when tumors with an unfavorable follow-up are selected for radical treatment, patients may comprise a selected group, with outcomes that may be different when compared with men who are operated immediately after diagnosis.
When compared with the available literature, Warlick et al performed a similar analysis, although in a smaller patient cohort.20 The outcomes of 38 men who had been included in a prospective expectant management protocol with T1c, PSA density <0.15 (still, 11 of 38 had a PSA density ≥0.15), ≤2 positive cores, no Gleason pattern 4 or 5, and <50% of any core involved with cancer-prostate cancer after delayed surgical intervention (median delay, 26.5 months) were compared with 150 matched patients who received immediate surgical intervention (median delay, 3.0 months). Outcome was defined as a <75% nomogram-derived chance of biochemical recurrence-free 10-year post-RP survival. The nomogram incorporated RP Gleason score, PSA, and RP organ-confinement status. A significant difference (23% of patients scored <75% in the delayed group vs 16% in the matched group) was not found.
Khatami et al performed an exploratory case-control analysis of 28 Swedish men in an active surveillance setting, who were also included in the current delayed RP group.21 Initial surveillance with radical treatment at the moment of progression did not seem to compromise curability. Patel et al also concluded that radical treatment at the time of progression after initial expectant management is effective.22
Freedland et al studied the association of time between diagnosis and surgery between 1988 and 2004 in a population within the SEARCH database of 895 men with a PSA <10 ng/mL diagnosed with prostate cancer, with a Gleason score of ≤6, dividing the interval between diagnosis and surgery into ≤90 days, 91 to 180 days, or >180 days.19 It was not reported whether this cohort concerns screen-detected or clinically detected patients, or whether the delay was intended, as in an active surveillance-like strategy. No significant differences were found in high-grade disease, positive surgical margins, or extraprostatic extension. Men with >180 days delay did, however, show an increased risk of biochemical progression (P = .002).
In a similar retrospective population setting, Nam et al found a possible relationship between a delay of >3 months and adverse outcomes of 645 RPs performed between 1987 and 1997.23 Finally, Khan et al found no impact on cancer control after a delay of >60 days of 926 RP's performed between 1989 and 1994.24
Weaknesses of our study include the lack of a standardized follow-up protocol during expectant management, such as used in current prospective active surveillance studies,11, 12 resulting in a situation in which specific reasons for men to switch to radical treatment were unknown. Second, the mean follow-up time was only 5.7 years, and only intermediate outcome variables instead of hard endpoints such as prostate cancer-specific mortality could be assessed. However, this endpoint may well never be reached in men with such a favorable form of prostate cancer.3 Finally, as stated above, this study design is surrogate for a randomized study in which RP would be performed after a fixed time interval in 1 group. A strength of our study is that it concerns a contemporary and representative cohort of patients within the controlled and standardized study environment of the ERSPC.
In the future, in addition to studying longer follow-up of retrospective data, it is essential to study the findings at and after RP in men who participate in current ongoing prospective active surveillance protocols that aim to filter out early (eg, not longer than 1-year delay) during follow-up the men who need active therapy, to assess the safety of these protocols.11, 12
In men with small, localized, well-differentiated, screen-detected prostate cancer, a statistically significant difference in the frequency of unfavorable histopathological and biochemical outcomes after RP was not found between men who received immediate surgery and men who received surgery after an initial period of expectant management after diagnosis. The time interval between diagnosis and RP was not significantly predictive for any adverse outcomes within these 2 groups. The harmful effect of treatment delay in prostate cancer patients eligible for active surveillance is therefore most likely to be limited. However, patients and physicians who choose expectant management as the initial treatment strategy should understand that if RP is indicated later based on follow-up, a selection bias may cause a higher frequency of unfavorable outcomes. At this moment, active surveillance is most appropriate for men older than 65 years at diagnosis; in younger men it should be recommended only to selected cases, as the long-term risk is unknown. Results of prospective active surveillance studies should be awaited.
CONFLICT OF INTEREST DISCLOSURES
The European Randomized Study of Screening for Prostate Cancer (ERSPC) is cofunded Europe-wide by Beckman Coulter Ltd.
In the Netherlands, the ERSPC is supported by grants from the Dutch Cancer Society (KWF 94-869, 98-1657, 2002-277, and 2006-3518), the Netherlands Organization for Health Research and Development (002822820, 22000106, and 50-50,110-98-311), sixth Framework Program of the EU (P-Mark: LSHC-CT-2004-503011), Beckman Coulter Hybritech Inc, and Europe Against Cancer (SOC 95 35,109, SOC 96 201,869 05F02, SOC 97 201329, and SOC 98 32,241). The ERSPC received Erasmus University Medical Center and Ministry of Health institutional review board approval. In Sweden, the ERSPC is supported by the Swedish Cancer Society (3792-B96-01XAB), Wallac Oy, Hybritech Inc, Schering-Plough Sweden, Abbot Pharmaceuticals Sweden, and grants from the Gunvor and Ivan Svensson Foundation and the Af Jochnick Foundation. Dr. Schröder is a consultant to Ferring Ltd, GlaxoSmith Kline, Bayer Schering, Cougar Biotechnology, and Genprobe.