- Email: [email protected]
Prognostic Markers Under Watchful Waiting and Radical Prostatectomy
Hematol Oncol Clin N Am 20 (2006) 845–855
HEMATOLOGY/ONCOLOGY CLINICS OF NORTH AMERICA
Prognostic Markers Under Watchful Waiting and Radical Prostatectomy Lars Holmberg, MD, PhDa,b,*, Anna Bill-Axelson, MD, PhDb, Hans Garmo, PhDa, Juni Palmgren, PhDc, Bo Johan Norle´n, MD, PhDb, Hans Olov Adami, MD, PhDc,d, Jan Erik Johansson, MD, PhDe, SPCG-4 Study Group a
Regional Oncologic Center, University Hospital, Uppsala, Sweden Department of Surgical Sciences, University Hospital, Uppsala, Sweden c Department of Medical Epidemiology and Biostatistics, Karolinska Institute, Stockholm, Sweden d Harvard School of Public Health, 677 Huntington Avenue, Boston, MA 02115, USA e ¨ rebro University Hospital, O ¨ rebro, Sweden Department of Urology, O b
C
onvincing empirical data show that radical treatment can alter the natural history of early, localized prostate cancer [1,2]. Nevertheless, the choice between surgery and watchful waiting will continue to be an issue in clinical decision making. Following information, some patients may decide that the risk of side effects of the treatment may not be worth taking in the light of how large a benefit he may anticipate. Thus, to aid decision making it is worth reviewing the prognosis by subgroups defined by patient characteristics at diagnosis. Until validated molecular signatures defined in diagnostic biopsies have been developed, such clinical patient characteristics must indeed be used in clinical routine. A suitable setting to analyze factors that determine prognosis or treatment response is an unbiased comparison of radical prostatectomy and watchful waiting as in the Scandinavian Prostate Cancer Group Trial number 4 (SPCG-4). In our previous presentation of 10-year results [1], we studied Gleason score, serum prostate specific antigen (PSA) at diagnosis, and age at diagnosis as modifiers of the effect of radical prostatectomy on survival. Because overall prognostic information obtained by these parameters or by tumor stage was not provided in our publication [1], we now present these data in the two study arms separately.
Experiences from the Scandinavian Prostate Cancer Group Trial No. 4.
*Corresponding author. Regional Oncologic Center, University Hospital, SE-751 85 Uppsala, Sweden. E-mail address: [email protected] (L. Holmberg). 0889-8588/06/$ – see front matter doi:10.1016/j.hoc.2006.03.007
ª 2006 Elsevier Inc. All rights reserved. hemonc.theclinics.com
846
HOLMBERG, BILL-AXELSON, GARMO, ET AL
SUBJECTS AND METHODS Details concerning study design and methods have been published previously [1,3]. The study protocol is available at www.roc.se. From 1989 to 1999, 695 men from 14 centers in Sweden, Finland, and Iceland were enrolled. The eligibility criteria included age younger than 75 years; the presence of a newly diagnosed, untreated localized prostate cancer; tumor stage of T0d (later classification T1b), T1 or T2 (T1c was included in 1994); a health status that would permit radical prostatectomy; and a life expectancy of more than 10 years. The tumor had to be well differentiated to moderately well differentiated, according to the definition established by the World Health Organization [4]. Patients had to have a bone scan that showed no abnormalities and a PSA level of less than 50 ng/mL. Patients were randomly assigned to undergo either radical prostatectomy or watchful waiting. Stratification was made according to tumor grade and randomization center. For men assigned to the radical prostatectomy group, surgery started with dissection of the pelvic lymph nodes. If there were no signs of metastases in frozen sections, the operation was continued with retropubic radical prostatectomy. The men in the watchful waiting group received no initial treatment other than the transurethral resection some of them had already undergone. Up to the end of 2003, 21 men assigned to radical prostatectomy had not undergone surgery, and 43 men assigned to watchful waiting had undergone treatment with curative intent. Hormonal treatment was recommended for men with symptomatic local progression in the radical prostatectomy group and for those with disseminated disease in both groups. In January 2003, an amendment to the protocol allowed men in both groups to begin hormonal therapy if their physicians advised it. Follow-up and Definition of Clinical Events The participants were seen every 6 months during the first 2 years and then annually for clinical examination and blood tests. A bone scan and a chest radiograph were obtained annually until 1997; thereafter, chest radiographs were obtained only once a year for the first 2 years after randomization. Beginning in 2003, bone scans were allowed every second year if the patient had no biochemical or clinical signs of progression. During 2004, all patient records were retrieved and individually reviewed for new events and for treatments started as a result of prostate cancer. An independent endpoint committee determined the cause of death on the basis of standardized extractions from the patient files and treatment group was not revealed to the committee. Distant metastases were considered present when bone scans, skeletal radiographs, computed tomographic scans, or chest radiographs revealed metastases or if lymph nodes at sites other than the regional sites showed cytologic or histologic evidence of prostate cancer. Statistical Analysis For the purpose of this analysis we used three main endpoints. First, we used disease-specific death, with death as a result of prostate cancer as specified in
WATCHFUL WAITING AND RADICAL PROSTATECTOMY
847
the main publication of the study [1]. In this analysis prostate cancer was considered the event and death from other causes treated as competing risks. Second, we analyzed time to distant metastases, with its diagnosis considered the event and death from other causes treated as competing risks. Third, we used time to a start of any systemic hormonal treatment, regardless of the indication for the treatment. Date of start of hormonal treatment was extracted from the review of all medical records that is continuously ongoing in the follow-up of the study. All analyses were performed in accordance with the intention-to-treat principle. The main measures used are cumulative incidence proportions stratified on Gleason grade, PSA level at diagnosis, tumor stage, and age group. We also estimated relative risks from Cox proportional hazards models where competing risk events were treated as censoring events. In our previous analysis [1] we tested any modification of the effect of radical prostatectomy according to subgroup with a Cox proportional hazards model, which included an interaction term between subgroup category and randomization group. The subgroups as explored in that analysis were Gleason score in the pre-randomization biopsy specimen less than 7 as compared with greater than or equal to 7, PSA level at diagnosis with 10 ng/mL or lower as compared with more than 10 ng/mL, and age younger than 65 years as compared with 65 years of age or older. In the present study we also analyzed tumor stage by comparing all T0 or T1 tumors to T2 tumors. The cut-offs to define the subgroups were defined on a theoretical basis before the data were explored. RESULTS Table 1 shows the distribution of patient characteristics and the number of events by randomization arm. The patient characteristics are similarly distributed between the radical prostatectomy and the watchful waiting groups. The prostate cancer–related events are markedly more common following watchful waiting than radical prostatectomy, as reported in detail previously [1,3]. Worsening of the prognosis was positively correlated with Gleason score, PSA at diagnosis, and tumor stage. As shown in our previous interaction analysis [1], younger age correlated to worse prognosis in the watchful waiting arm, but not in the radical prostatectomy arm (Table 2). In the watchful waiting arm the two subgroups with the lowest 10-year cumulative incidences of death resulting from prostate cancer were those defined by Gleason score less than 7 and by PSA less than or equal to 10 ng/mL at diagnosis. In the watchful waiting arm, only men with Gleason score less than 7 had a risk of prostate cancer death under 1% per year (Table 2). Based on the findings in Table 2, we defined a posteriori group of men with a Gleason score less than 7 and 65 years of age and older. Among these men, the cumulative incidence of prostate cancer death at 10 years was 11%. In all subgroups—with the exception of men 65 years of age and older—the difference in cumulative incidence of death from prostate cancer between the watchful waiting and radical prostatectomy groups was at least 3% or higher; the estimates varied between 3.9% and 12.9%.
848
HOLMBERG, BILL-AXELSON, GARMO, ET AL
Table 1 Patient characteristics and number of events by randomization arm Radical prostatectomy (n ¼ 347) Age mean, y (SE) PSA mean, ng/mL (SE) Tumor stage (%)a T1b T1c T2 Unknown Gleason score (%)b Gleason 2–4 Gleason 5–6 Gleason 7 Gleason 8–10 Unknownc PSA, ng/mL (%) <4 4–6.9 7–10 10.1–20 >20 Unknown Status at end of follow-up (%) Alive PC-death Death from other cause Other main cause, but with metastases Other main cause, without metastases but with local progression/recurrence Other main cause, and no evidence of metastases or local progression/ recurrence Other main cause, within first month after randomization Metastasis staus at end of follow-up (%) Metastases free Metastasis Death without metastasis Mean follow-up time (SE)
64.6 13.5
(5.1) (10.7)
Watchful waiting (n ¼ 348) 64.5 12.3
(5.0) (10.3)
33 43 270 1
(9.5) (12.4) (77.8) (0.3)
50 38 259 1
(14.4) (10.9) (74.4) (0.3)
44 165 77 14 47
(12.7) (47.6) (22.2) (4.0) (13.5)
46 166 82 21 33
(13.2) (47.7) (23.6) (6.0) (9.5)
43 60 68 100 69 7
(12.4) (17.3) (19.6) (28.8) (19.9) (2.0)
63 60 67 95 60 3
(18.1) (17.2) (19.3) (27.3) (17.2) (0.9)
264 30 53 1 6
(76.1) (8.6) (15.3) (0.3) (1.7)
242 50 56 8 13
(69.5) (14.4) (16.1) (2.3) (3.7)
45
(13.0)
34
(9.8)
1
(0.3)
1
245 50 52 8.8
(0.3)
(70.6) (14.4) (15.0) (2.8)
221 79 48 8.5
(63.5) (22.7) (13.8) (2.8)
Abbreviations: PC, prostate cancer; PSA, prostate specific antigen. a T1b indicates an incidental histologic finding in more than 5% of tissue resected (in 1978, this was classified as stage T0d); stage T1c indicates a tumor identified by needle biopsy because of elevated serum prostate specific antigen levels (in 1978 this classification did not exist). In palpable carcinoma confined to the prostate, T2 indicates a tumor confined within the prostate (in 1978, this was classified as T1 or T2). b This score was assigned during histopathological review. c Diagnosis was made by cytologic examination only in 55 patients; a biopsy specimen could not be retrieved in 24 patients.
WATCHFUL WAITING AND RADICAL PROSTATECTOMY
849
Table 2 Cumulative incidence at 10 years (percent) and 95% confidence limits of prostate cancer death by subgroup and randomization arm
Gleason < 7 Gleason 7 PSA 10 PSA > 10 T1b þ T1c T2 Age < 65 Age 65
Radical prostatectomy
Watchful waiting
3.5 15.2 5.5 14.4 5.2 10.4 8.5 10.7
7.4 28.1 10.0 20.4 10.9 16.2 19.2 11.0
(1.5, (8.6, (2.6, (9.1, (1.7, (6.9, (4.7, (6.3,
7.7) 26.8) 11.8) 22.7) 16.2) 15.7) 15.4) 18.0)
(4.2, 3.0) (19.7, 40.2) (6.1, 16.5) (14.4, 29.0) (4.8, 24.6) (12.0, 21.9) (13.5, 27.2) (6.8, 17.8)
Abbreviation: PSA, prostate specific antigen.
Because the development over time is of interest, the findings in Table 2 are additionally illustrated in Figs. 1–4. In the subgroups with a better prognosis (low Gleason score, low PSA at diagnosis, and lesser tumor stage), the difference between radical prostatectomy and watchful waiting emerges first 5 to 6 years after diagnosis. Looking at the relative hazards associated with having a Gleason score of 7 and over or a PSA more than 10 ng/mL, the relative hazards were statistically significant over unity in the range of 2.27–3.97 (Table 3). In the strata defined by tumor stage and age, the relative hazards were lower and the confidence intervals included 1.0. Thus, Gleason score and PSA level at diagnosis were the strongest risk factors and also behaved consistently in both study arms, while age only was weakly correlated to prognosis and showed a pattern of interaction with randomization arm (Table 3) [1]. To further explore if men with tumors with Gleason score under 7 should be recommended watchful waiting, we also studied the cumulative incidence of distant metastases. However, after 4 years, the cumulative incidence curve in the watchful waiting arm began to separate from the curve in the radical prostatectomy group and reached 17% after 10 years of observation. When we analyzed the cumulative incidence of hormonal treatment in the two study arms as an indicator of failure of the initial treatment strategy, the watchful waiting arm reached 57% after 10 years and the radical prostatectomy arm 34%. In all subgroups of the watchful waiting arm, the cumulative incidence exceeded 40% at 10 years, even so in the group with Gleason score below 7 (Fig. 5). SUMMARY In accordance with most other studies, we found a clear association between Gleason score and PSA level at diagnosis on one hand and prognosis on the other. Tumor stage and age at randomization were not as strongly related to risk of prostate cancer death. The data presented here—and an interaction analysis in our main presentation of the study [1]—show that there is no clear
HOLMBERG, BILL-AXELSON, GARMO, ET AL
850
A
Cumulative incidence of PC-death Gleason < 7
Probability
0.4 0.3
Radical prostatectomy Watchful waiting
0.2 0.1 0 0
2
4
6
8
10
172 179
120 127
65 70
Years No. At Risk Radical prostatectomy Watchful waiting
209 212
B
207 209
204 202
Cumulative incidence of PC-death Gleason >= 7
Probability
0.4 0.3
Radical prostatectomy Watchful waiting
0.2 0.1 0 0
2
4
6
8
10
71 76
55 50
32 24
Years No. At Risk Radical prostatectomy Watchful waiting
91 103
90 102
82 96
Fig. 1. Cumulative incidence of prostate cancer death by Gleason score and randomization arm. (A) Patients with Gleason score less than 7. (B) Gleason score 7 and over.
evidence that Gleason score, PSA level at diagnosis, or tumor stage modifies the beneficial effect of radical prostatectomy; in all these subgroups there is a benefit from surgery. However, in the analysis by age group, there was a similar cumulative incidence curve in both age groups following randomization to prostatectomy and for the older men also following allocation to watchful waiting. In contrast, the prognosis in the youngest age group in the watchful waiting group was considerably worse resulting in the impression of an effect modification on radical prostatectomy by age [1]. After 10 years of followup, the probability to be on hormonal treatment among the men in the watchful waiting group is close to 50%. Subgroup analyses should always be interpreted cautiously. Statistical precision is diminished and the confidence intervals become wide (Table 1). Thus,
WATCHFUL WAITING AND RADICAL PROSTATECTOMY
A
851
Cumulative incidence of PC-death PSA <= 10
Probability
0.4 0.3
Radical prostatectomy Watchful waiting
0.2 0.1 0 0
2
4
6
8
10
142 156
109 115
62 60
Years No. At Risk Radical prostatectomy Watchful waiting
171 190
169 187
164 180
Cumulative incidence of PC-death PSA > 10
B
Probability
0.4 0.3
Radical prostatectomy Watchful waiting
0.2 0.1 0 0
2
4
6
8
10
136 120
96 80
52 42
Years No. At Risk Radical prostatectomy Watchful waiting
169 155
168 151
162 143
Fig. 2. Cumulative incidence of prostate cancer death by PSA at randomization. (A) Patients with PSA less than or equal to 10 ng/mL. (B) Patients with PSA more than 10 ng/mL.
care should be taken not to transfer the point estimates of this study directly to clinical decision; there may indeed be a broad range of values in each subgroup that are compatible with the real risks of prostate cancer death or distant metastases. The lack of statistical precision also implies a low power to detect a difference, should there be one. We therefore caution that our finding of a low or no benefit from radical surgery in the age group above 65 years should be taken mainly as a hypothesis worth testing in other settings. There is also a risk of spurious findings as a result of multiple comparisons. However in the present analysis, we have not relied on multiple testing procedures to find eventual statistically significant differences, but rather looked at overall patterns of prognosis. Our findings are in line with what can be expected given
HOLMBERG, BILL-AXELSON, GARMO, ET AL
852
Cumulative incidence of PC-death T-stage: T1b+T1c
A 0.4
Probability
0.3
Radical pros tatectom y Watchful waiting
0.2 0.1 0 0
2
4
6
8
10
61 68
41 44
16 22
Years No. At Risk Radical prostatectomy Watchful waiting
76 88
76 88
75 84
Cumulative incidence of PC-death T-stage: T2
B 0.4
Probability
0.3
Radical prostatectomy Watchful waiting
0.2
0.1
0 0
2
4
6
8
10
222 210
168 154
101 82
Years No. At Risk Radical prostatectomy Watchful waiting
270 259
266 252
256 241
Fig. 3. Cumulative incidence of prostate cancer death by T stage. (A) Patients with T0 and T1 tumors. (B) Patients with T2 tumors.
the data in other studies [4–6]. Furthermore, with one exception, our subgroups were defined a priori, and are thus not data-derived. We analyzed our data according to intention to treat. Thus, the results mirror the actual clinical decision-making procedure; they reflect the prognosis following an initial recommendation to wait and see or to operate. If patient compliance to randomization is low, any true difference between the interventions would be diluted. However, the crossover between study arms was limited in our study. A per protocol analysis would have influenced the difference between the study arms, but not the prognostic power of the patient
WATCHFUL WAITING AND RADICAL PROSTATECTOMY
853
Cumulative incidence of PC-death Age < 65
A 0.4
Probability
0.3
Radical prostatectomy Watchful waiting
0.2 0.1 0 0
2
4
6
8
10
135 128
110 96
66 56
Years No. At Risk Radical prostatectomy Watchful waiting
157 166
156 161
151 153
Cumulative incidence of PC-death Age >= 65
B 0.4
Probability
0.3
Radical prostatectomy Watchful waiting
0.2
0.1
0 0
2
4
6
8
10
149 151
100 102
52 48
Years No. At Risk Radical prostatectomy Watchful waiting
190 182
187 180
181 173
Fig. 4. Cumulative incidence by age group. (A) Patients with age younger than 65. (B) Patients with age 65 or older.
characteristics. The intention to treat approach also keeps the randomization principle to obtain an unbiased comparison intact throughout the analyses. The majority of men enrolled in the SPCG-4 trial had a clinically detected prostate cancer. This is different from the current situation in many parts of the world where PSA screening is common. That screening today is widespread has at least two important consequences for the generalizability of our findings. First, screening introduces a lead time, which may be between 5 and 12 years [7,8]. By definition, no clinical event could occur during the lead time. Thus, the time to reach the risk levels for progression and diseasespecific death observed in the SPCG-4 study would be delayed by the same magnitude of time. Furthermore, the age limit of 65 may not be the relevant
HOLMBERG, BILL-AXELSON, GARMO, ET AL
854
Table 3 Hazard ratios and 95% confidence limits from Cox models of risk of prostate cancer death within categories of Gleason score, PSA at diagnosis, tumor stage and age by randomization group
Gleason < 7 vs Gleason 7 PSA 10 vs PSA > 10 T1b þ T1c vs T2 Age 65 vs Age < 65
Radical prostatectomy
Watchful waiting
3.97 3.68 2.27 0.84
3.30 2.26 1.91 1.68
(1.65, (1.58, (0.69, (0.41,
9.60) 8.59) 7.49) 1.73)
(1.79, (1.25, (0.86, (0.94,
4.00) 2.69) 2.65) 1.98)
Abbreviation: PSA, prostate specific antigen.
one in a screening setting since the age at diagnosis at a mean also will be advanced by the length of the lead time. Second, intense PSA screening using low PSA thresholds for biopsy may lead to detection of biologically indolent or slow-growing tumors that generally are small and well-differentiated. In studies comprising chiefly PSA-detected prostate cancer, the distribution of Gleason score, level of PSA at diagnosis, and tumor stage may also be so limited that their prognostic value becomes difficult to ascertain. In such studies, the associations between these factors and prognosis might also be different than in studies with a high proportion of men with clinically significant prostate cancer, such as the SPCG-4 trial. Cumulative incidence of hormonal treatment Gleason < 7 0.8
Radical prostatectomy Watchful waiting
Probability
0.6
0.4
0.2
0 0
2
4
6
8
10
142 131
94 79
42 40
Years No. At Risk Radical prostatectomy Watchful waiting
209 212
198 194
184 174
Fig. 5. Cumulative incidence of hormonal treatment by randomization arm in the subgroup defined by Gleason score less than 7.
WATCHFUL WAITING AND RADICAL PROSTATECTOMY
855
If we regard the start of hormonal treatment as a proxy for failure, the risk of failure of any kind is high in the watchful waiting group. The indications for treatment are a mixture of rising PSA levels, emerging local symptoms, verified local progression, and appearance of distant metastases. Probably, also factors such as patients’ and doctors’ anxiety play a role. With the possible exception of age older than 65 at diagnosis, we were not able to define any subgroup of men—in our trial comprising mainly clinically detected prostate cancer—in which watchful waiting is better than radical prostatectomy in terms of protection from disease progression. However, the analyses also emphasize the need for more empirical data to guide clinical decision and patient counseling. Until then, we need to acknowledge that in some prognostic subgroups risks are so low also following watchful waiting that the decision to treat needs to be carefully weighed against the different patterns of side effects [9,10]. In this study, a subdivision by the Gleason score seems to be the best candidate to separate low- and high-risk groups, thus pointing further to the hope that more sophisticated biological markers in the tumor cells will be of great help in future clinical decision making. Acknowledgment The study was supported by the Swedish Cancer Society. References [1] Bill-Axelson A, Holmberg L, Ruutu M, et al. Radcial prostatectomy versus watchful waiting in early prostate cancer. N Engl J Med 2005;352:1977–84. [2] Jani AB, Hellman S. Early prostate cancer: clinical decision-making. Lancet 2003;361: 1045–53. [3] Holmberg L, Bill-Axelson A, Helgesen F, et al. A randomized trial comparing radical prostatectomy with watchful waiting in early prostate cancer. N Engl J Med 2002;347:781–9. [4] Mostofi FK, Sesterhenn I, Sobin LH. Histological typing of prostate tumors. In: International classification of tumors. WHO monograph series no. 22. Geneva: World Health Organization; 1980. [5] Johansson JE, Andren O, Andersson SO, et al. Natural history of early localized prostate cancer. JAMA 2004;291:2713–9. [6] Zietman AL, Thakral H, Wilson L, et al. Conservative management of prostate cancer in the prostate specific antigen era: the incidence and time course of subsequent therapy. J Urol 2001;166:1702–6. [7] Soloway M, Roach M. Prostate cancer progression after therapy of primary curative intent. A review of data from the prostate-specific antigen era. Cancer 2005;104:2310–22. [8] To ¨ rnblom M, Eriksson H, Franze´n S, et al. Lead time associated with screening for prostate cancer. Int J Cancer 2004;108:122–9. [9] Draisma G, Boer R, Otto SJ, et al. Lead times and overdetection due to prostate-specific antigen screening: estimates from the European Randomized Study of Screening for Prostate Cancer. J Natl Cancer Inst 2003;95:868–78. [10] Steineck G, Helgesen F, Adolfsson J, et al. Scandinavian Prostate Cancer Group Study number 4. Quality of life after radical prostatectomy or watchful waiting. N Engl J Med 2002;347:790–6.
HEMATOLOGY/ONCOLOGY CLINICS OF NORTH AMERICA
Prognostic Markers Under Watchful Waiting and Radical Prostatectomy Lars Holmberg, MD, PhDa,b,*, Anna Bill-Axelson, MD, PhDb, Hans Garmo, PhDa, Juni Palmgren, PhDc, Bo Johan Norle´n, MD, PhDb, Hans Olov Adami, MD, PhDc,d, Jan Erik Johansson, MD, PhDe, SPCG-4 Study Group a
Regional Oncologic Center, University Hospital, Uppsala, Sweden Department of Surgical Sciences, University Hospital, Uppsala, Sweden c Department of Medical Epidemiology and Biostatistics, Karolinska Institute, Stockholm, Sweden d Harvard School of Public Health, 677 Huntington Avenue, Boston, MA 02115, USA e ¨ rebro University Hospital, O ¨ rebro, Sweden Department of Urology, O b
C
onvincing empirical data show that radical treatment can alter the natural history of early, localized prostate cancer [1,2]. Nevertheless, the choice between surgery and watchful waiting will continue to be an issue in clinical decision making. Following information, some patients may decide that the risk of side effects of the treatment may not be worth taking in the light of how large a benefit he may anticipate. Thus, to aid decision making it is worth reviewing the prognosis by subgroups defined by patient characteristics at diagnosis. Until validated molecular signatures defined in diagnostic biopsies have been developed, such clinical patient characteristics must indeed be used in clinical routine. A suitable setting to analyze factors that determine prognosis or treatment response is an unbiased comparison of radical prostatectomy and watchful waiting as in the Scandinavian Prostate Cancer Group Trial number 4 (SPCG-4). In our previous presentation of 10-year results [1], we studied Gleason score, serum prostate specific antigen (PSA) at diagnosis, and age at diagnosis as modifiers of the effect of radical prostatectomy on survival. Because overall prognostic information obtained by these parameters or by tumor stage was not provided in our publication [1], we now present these data in the two study arms separately.
Experiences from the Scandinavian Prostate Cancer Group Trial No. 4.
*Corresponding author. Regional Oncologic Center, University Hospital, SE-751 85 Uppsala, Sweden. E-mail address: [email protected] (L. Holmberg). 0889-8588/06/$ – see front matter doi:10.1016/j.hoc.2006.03.007
ª 2006 Elsevier Inc. All rights reserved. hemonc.theclinics.com
846
HOLMBERG, BILL-AXELSON, GARMO, ET AL
SUBJECTS AND METHODS Details concerning study design and methods have been published previously [1,3]. The study protocol is available at www.roc.se. From 1989 to 1999, 695 men from 14 centers in Sweden, Finland, and Iceland were enrolled. The eligibility criteria included age younger than 75 years; the presence of a newly diagnosed, untreated localized prostate cancer; tumor stage of T0d (later classification T1b), T1 or T2 (T1c was included in 1994); a health status that would permit radical prostatectomy; and a life expectancy of more than 10 years. The tumor had to be well differentiated to moderately well differentiated, according to the definition established by the World Health Organization [4]. Patients had to have a bone scan that showed no abnormalities and a PSA level of less than 50 ng/mL. Patients were randomly assigned to undergo either radical prostatectomy or watchful waiting. Stratification was made according to tumor grade and randomization center. For men assigned to the radical prostatectomy group, surgery started with dissection of the pelvic lymph nodes. If there were no signs of metastases in frozen sections, the operation was continued with retropubic radical prostatectomy. The men in the watchful waiting group received no initial treatment other than the transurethral resection some of them had already undergone. Up to the end of 2003, 21 men assigned to radical prostatectomy had not undergone surgery, and 43 men assigned to watchful waiting had undergone treatment with curative intent. Hormonal treatment was recommended for men with symptomatic local progression in the radical prostatectomy group and for those with disseminated disease in both groups. In January 2003, an amendment to the protocol allowed men in both groups to begin hormonal therapy if their physicians advised it. Follow-up and Definition of Clinical Events The participants were seen every 6 months during the first 2 years and then annually for clinical examination and blood tests. A bone scan and a chest radiograph were obtained annually until 1997; thereafter, chest radiographs were obtained only once a year for the first 2 years after randomization. Beginning in 2003, bone scans were allowed every second year if the patient had no biochemical or clinical signs of progression. During 2004, all patient records were retrieved and individually reviewed for new events and for treatments started as a result of prostate cancer. An independent endpoint committee determined the cause of death on the basis of standardized extractions from the patient files and treatment group was not revealed to the committee. Distant metastases were considered present when bone scans, skeletal radiographs, computed tomographic scans, or chest radiographs revealed metastases or if lymph nodes at sites other than the regional sites showed cytologic or histologic evidence of prostate cancer. Statistical Analysis For the purpose of this analysis we used three main endpoints. First, we used disease-specific death, with death as a result of prostate cancer as specified in
WATCHFUL WAITING AND RADICAL PROSTATECTOMY
847
the main publication of the study [1]. In this analysis prostate cancer was considered the event and death from other causes treated as competing risks. Second, we analyzed time to distant metastases, with its diagnosis considered the event and death from other causes treated as competing risks. Third, we used time to a start of any systemic hormonal treatment, regardless of the indication for the treatment. Date of start of hormonal treatment was extracted from the review of all medical records that is continuously ongoing in the follow-up of the study. All analyses were performed in accordance with the intention-to-treat principle. The main measures used are cumulative incidence proportions stratified on Gleason grade, PSA level at diagnosis, tumor stage, and age group. We also estimated relative risks from Cox proportional hazards models where competing risk events were treated as censoring events. In our previous analysis [1] we tested any modification of the effect of radical prostatectomy according to subgroup with a Cox proportional hazards model, which included an interaction term between subgroup category and randomization group. The subgroups as explored in that analysis were Gleason score in the pre-randomization biopsy specimen less than 7 as compared with greater than or equal to 7, PSA level at diagnosis with 10 ng/mL or lower as compared with more than 10 ng/mL, and age younger than 65 years as compared with 65 years of age or older. In the present study we also analyzed tumor stage by comparing all T0 or T1 tumors to T2 tumors. The cut-offs to define the subgroups were defined on a theoretical basis before the data were explored. RESULTS Table 1 shows the distribution of patient characteristics and the number of events by randomization arm. The patient characteristics are similarly distributed between the radical prostatectomy and the watchful waiting groups. The prostate cancer–related events are markedly more common following watchful waiting than radical prostatectomy, as reported in detail previously [1,3]. Worsening of the prognosis was positively correlated with Gleason score, PSA at diagnosis, and tumor stage. As shown in our previous interaction analysis [1], younger age correlated to worse prognosis in the watchful waiting arm, but not in the radical prostatectomy arm (Table 2). In the watchful waiting arm the two subgroups with the lowest 10-year cumulative incidences of death resulting from prostate cancer were those defined by Gleason score less than 7 and by PSA less than or equal to 10 ng/mL at diagnosis. In the watchful waiting arm, only men with Gleason score less than 7 had a risk of prostate cancer death under 1% per year (Table 2). Based on the findings in Table 2, we defined a posteriori group of men with a Gleason score less than 7 and 65 years of age and older. Among these men, the cumulative incidence of prostate cancer death at 10 years was 11%. In all subgroups—with the exception of men 65 years of age and older—the difference in cumulative incidence of death from prostate cancer between the watchful waiting and radical prostatectomy groups was at least 3% or higher; the estimates varied between 3.9% and 12.9%.
848
HOLMBERG, BILL-AXELSON, GARMO, ET AL
Table 1 Patient characteristics and number of events by randomization arm Radical prostatectomy (n ¼ 347) Age mean, y (SE) PSA mean, ng/mL (SE) Tumor stage (%)a T1b T1c T2 Unknown Gleason score (%)b Gleason 2–4 Gleason 5–6 Gleason 7 Gleason 8–10 Unknownc PSA, ng/mL (%) <4 4–6.9 7–10 10.1–20 >20 Unknown Status at end of follow-up (%) Alive PC-death Death from other cause Other main cause, but with metastases Other main cause, without metastases but with local progression/recurrence Other main cause, and no evidence of metastases or local progression/ recurrence Other main cause, within first month after randomization Metastasis staus at end of follow-up (%) Metastases free Metastasis Death without metastasis Mean follow-up time (SE)
64.6 13.5
(5.1) (10.7)
Watchful waiting (n ¼ 348) 64.5 12.3
(5.0) (10.3)
33 43 270 1
(9.5) (12.4) (77.8) (0.3)
50 38 259 1
(14.4) (10.9) (74.4) (0.3)
44 165 77 14 47
(12.7) (47.6) (22.2) (4.0) (13.5)
46 166 82 21 33
(13.2) (47.7) (23.6) (6.0) (9.5)
43 60 68 100 69 7
(12.4) (17.3) (19.6) (28.8) (19.9) (2.0)
63 60 67 95 60 3
(18.1) (17.2) (19.3) (27.3) (17.2) (0.9)
264 30 53 1 6
(76.1) (8.6) (15.3) (0.3) (1.7)
242 50 56 8 13
(69.5) (14.4) (16.1) (2.3) (3.7)
45
(13.0)
34
(9.8)
1
(0.3)
1
245 50 52 8.8
(0.3)
(70.6) (14.4) (15.0) (2.8)
221 79 48 8.5
(63.5) (22.7) (13.8) (2.8)
Abbreviations: PC, prostate cancer; PSA, prostate specific antigen. a T1b indicates an incidental histologic finding in more than 5% of tissue resected (in 1978, this was classified as stage T0d); stage T1c indicates a tumor identified by needle biopsy because of elevated serum prostate specific antigen levels (in 1978 this classification did not exist). In palpable carcinoma confined to the prostate, T2 indicates a tumor confined within the prostate (in 1978, this was classified as T1 or T2). b This score was assigned during histopathological review. c Diagnosis was made by cytologic examination only in 55 patients; a biopsy specimen could not be retrieved in 24 patients.
WATCHFUL WAITING AND RADICAL PROSTATECTOMY
849
Table 2 Cumulative incidence at 10 years (percent) and 95% confidence limits of prostate cancer death by subgroup and randomization arm
Gleason < 7 Gleason 7 PSA 10 PSA > 10 T1b þ T1c T2 Age < 65 Age 65
Radical prostatectomy
Watchful waiting
3.5 15.2 5.5 14.4 5.2 10.4 8.5 10.7
7.4 28.1 10.0 20.4 10.9 16.2 19.2 11.0
(1.5, (8.6, (2.6, (9.1, (1.7, (6.9, (4.7, (6.3,
7.7) 26.8) 11.8) 22.7) 16.2) 15.7) 15.4) 18.0)
(4.2, 3.0) (19.7, 40.2) (6.1, 16.5) (14.4, 29.0) (4.8, 24.6) (12.0, 21.9) (13.5, 27.2) (6.8, 17.8)
Abbreviation: PSA, prostate specific antigen.
Because the development over time is of interest, the findings in Table 2 are additionally illustrated in Figs. 1–4. In the subgroups with a better prognosis (low Gleason score, low PSA at diagnosis, and lesser tumor stage), the difference between radical prostatectomy and watchful waiting emerges first 5 to 6 years after diagnosis. Looking at the relative hazards associated with having a Gleason score of 7 and over or a PSA more than 10 ng/mL, the relative hazards were statistically significant over unity in the range of 2.27–3.97 (Table 3). In the strata defined by tumor stage and age, the relative hazards were lower and the confidence intervals included 1.0. Thus, Gleason score and PSA level at diagnosis were the strongest risk factors and also behaved consistently in both study arms, while age only was weakly correlated to prognosis and showed a pattern of interaction with randomization arm (Table 3) [1]. To further explore if men with tumors with Gleason score under 7 should be recommended watchful waiting, we also studied the cumulative incidence of distant metastases. However, after 4 years, the cumulative incidence curve in the watchful waiting arm began to separate from the curve in the radical prostatectomy group and reached 17% after 10 years of observation. When we analyzed the cumulative incidence of hormonal treatment in the two study arms as an indicator of failure of the initial treatment strategy, the watchful waiting arm reached 57% after 10 years and the radical prostatectomy arm 34%. In all subgroups of the watchful waiting arm, the cumulative incidence exceeded 40% at 10 years, even so in the group with Gleason score below 7 (Fig. 5). SUMMARY In accordance with most other studies, we found a clear association between Gleason score and PSA level at diagnosis on one hand and prognosis on the other. Tumor stage and age at randomization were not as strongly related to risk of prostate cancer death. The data presented here—and an interaction analysis in our main presentation of the study [1]—show that there is no clear
HOLMBERG, BILL-AXELSON, GARMO, ET AL
850
A
Cumulative incidence of PC-death Gleason < 7
Probability
0.4 0.3
Radical prostatectomy Watchful waiting
0.2 0.1 0 0
2
4
6
8
10
172 179
120 127
65 70
Years No. At Risk Radical prostatectomy Watchful waiting
209 212
B
207 209
204 202
Cumulative incidence of PC-death Gleason >= 7
Probability
0.4 0.3
Radical prostatectomy Watchful waiting
0.2 0.1 0 0
2
4
6
8
10
71 76
55 50
32 24
Years No. At Risk Radical prostatectomy Watchful waiting
91 103
90 102
82 96
Fig. 1. Cumulative incidence of prostate cancer death by Gleason score and randomization arm. (A) Patients with Gleason score less than 7. (B) Gleason score 7 and over.
evidence that Gleason score, PSA level at diagnosis, or tumor stage modifies the beneficial effect of radical prostatectomy; in all these subgroups there is a benefit from surgery. However, in the analysis by age group, there was a similar cumulative incidence curve in both age groups following randomization to prostatectomy and for the older men also following allocation to watchful waiting. In contrast, the prognosis in the youngest age group in the watchful waiting group was considerably worse resulting in the impression of an effect modification on radical prostatectomy by age [1]. After 10 years of followup, the probability to be on hormonal treatment among the men in the watchful waiting group is close to 50%. Subgroup analyses should always be interpreted cautiously. Statistical precision is diminished and the confidence intervals become wide (Table 1). Thus,
WATCHFUL WAITING AND RADICAL PROSTATECTOMY
A
851
Cumulative incidence of PC-death PSA <= 10
Probability
0.4 0.3
Radical prostatectomy Watchful waiting
0.2 0.1 0 0
2
4
6
8
10
142 156
109 115
62 60
Years No. At Risk Radical prostatectomy Watchful waiting
171 190
169 187
164 180
Cumulative incidence of PC-death PSA > 10
B
Probability
0.4 0.3
Radical prostatectomy Watchful waiting
0.2 0.1 0 0
2
4
6
8
10
136 120
96 80
52 42
Years No. At Risk Radical prostatectomy Watchful waiting
169 155
168 151
162 143
Fig. 2. Cumulative incidence of prostate cancer death by PSA at randomization. (A) Patients with PSA less than or equal to 10 ng/mL. (B) Patients with PSA more than 10 ng/mL.
care should be taken not to transfer the point estimates of this study directly to clinical decision; there may indeed be a broad range of values in each subgroup that are compatible with the real risks of prostate cancer death or distant metastases. The lack of statistical precision also implies a low power to detect a difference, should there be one. We therefore caution that our finding of a low or no benefit from radical surgery in the age group above 65 years should be taken mainly as a hypothesis worth testing in other settings. There is also a risk of spurious findings as a result of multiple comparisons. However in the present analysis, we have not relied on multiple testing procedures to find eventual statistically significant differences, but rather looked at overall patterns of prognosis. Our findings are in line with what can be expected given
HOLMBERG, BILL-AXELSON, GARMO, ET AL
852
Cumulative incidence of PC-death T-stage: T1b+T1c
A 0.4
Probability
0.3
Radical pros tatectom y Watchful waiting
0.2 0.1 0 0
2
4
6
8
10
61 68
41 44
16 22
Years No. At Risk Radical prostatectomy Watchful waiting
76 88
76 88
75 84
Cumulative incidence of PC-death T-stage: T2
B 0.4
Probability
0.3
Radical prostatectomy Watchful waiting
0.2
0.1
0 0
2
4
6
8
10
222 210
168 154
101 82
Years No. At Risk Radical prostatectomy Watchful waiting
270 259
266 252
256 241
Fig. 3. Cumulative incidence of prostate cancer death by T stage. (A) Patients with T0 and T1 tumors. (B) Patients with T2 tumors.
the data in other studies [4–6]. Furthermore, with one exception, our subgroups were defined a priori, and are thus not data-derived. We analyzed our data according to intention to treat. Thus, the results mirror the actual clinical decision-making procedure; they reflect the prognosis following an initial recommendation to wait and see or to operate. If patient compliance to randomization is low, any true difference between the interventions would be diluted. However, the crossover between study arms was limited in our study. A per protocol analysis would have influenced the difference between the study arms, but not the prognostic power of the patient
WATCHFUL WAITING AND RADICAL PROSTATECTOMY
853
Cumulative incidence of PC-death Age < 65
A 0.4
Probability
0.3
Radical prostatectomy Watchful waiting
0.2 0.1 0 0
2
4
6
8
10
135 128
110 96
66 56
Years No. At Risk Radical prostatectomy Watchful waiting
157 166
156 161
151 153
Cumulative incidence of PC-death Age >= 65
B 0.4
Probability
0.3
Radical prostatectomy Watchful waiting
0.2
0.1
0 0
2
4
6
8
10
149 151
100 102
52 48
Years No. At Risk Radical prostatectomy Watchful waiting
190 182
187 180
181 173
Fig. 4. Cumulative incidence by age group. (A) Patients with age younger than 65. (B) Patients with age 65 or older.
characteristics. The intention to treat approach also keeps the randomization principle to obtain an unbiased comparison intact throughout the analyses. The majority of men enrolled in the SPCG-4 trial had a clinically detected prostate cancer. This is different from the current situation in many parts of the world where PSA screening is common. That screening today is widespread has at least two important consequences for the generalizability of our findings. First, screening introduces a lead time, which may be between 5 and 12 years [7,8]. By definition, no clinical event could occur during the lead time. Thus, the time to reach the risk levels for progression and diseasespecific death observed in the SPCG-4 study would be delayed by the same magnitude of time. Furthermore, the age limit of 65 may not be the relevant
HOLMBERG, BILL-AXELSON, GARMO, ET AL
854
Table 3 Hazard ratios and 95% confidence limits from Cox models of risk of prostate cancer death within categories of Gleason score, PSA at diagnosis, tumor stage and age by randomization group
Gleason < 7 vs Gleason 7 PSA 10 vs PSA > 10 T1b þ T1c vs T2 Age 65 vs Age < 65
Radical prostatectomy
Watchful waiting
3.97 3.68 2.27 0.84
3.30 2.26 1.91 1.68
(1.65, (1.58, (0.69, (0.41,
9.60) 8.59) 7.49) 1.73)
(1.79, (1.25, (0.86, (0.94,
4.00) 2.69) 2.65) 1.98)
Abbreviation: PSA, prostate specific antigen.
one in a screening setting since the age at diagnosis at a mean also will be advanced by the length of the lead time. Second, intense PSA screening using low PSA thresholds for biopsy may lead to detection of biologically indolent or slow-growing tumors that generally are small and well-differentiated. In studies comprising chiefly PSA-detected prostate cancer, the distribution of Gleason score, level of PSA at diagnosis, and tumor stage may also be so limited that their prognostic value becomes difficult to ascertain. In such studies, the associations between these factors and prognosis might also be different than in studies with a high proportion of men with clinically significant prostate cancer, such as the SPCG-4 trial. Cumulative incidence of hormonal treatment Gleason < 7 0.8
Radical prostatectomy Watchful waiting
Probability
0.6
0.4
0.2
0 0
2
4
6
8
10
142 131
94 79
42 40
Years No. At Risk Radical prostatectomy Watchful waiting
209 212
198 194
184 174
Fig. 5. Cumulative incidence of hormonal treatment by randomization arm in the subgroup defined by Gleason score less than 7.
WATCHFUL WAITING AND RADICAL PROSTATECTOMY
855
If we regard the start of hormonal treatment as a proxy for failure, the risk of failure of any kind is high in the watchful waiting group. The indications for treatment are a mixture of rising PSA levels, emerging local symptoms, verified local progression, and appearance of distant metastases. Probably, also factors such as patients’ and doctors’ anxiety play a role. With the possible exception of age older than 65 at diagnosis, we were not able to define any subgroup of men—in our trial comprising mainly clinically detected prostate cancer—in which watchful waiting is better than radical prostatectomy in terms of protection from disease progression. However, the analyses also emphasize the need for more empirical data to guide clinical decision and patient counseling. Until then, we need to acknowledge that in some prognostic subgroups risks are so low also following watchful waiting that the decision to treat needs to be carefully weighed against the different patterns of side effects [9,10]. In this study, a subdivision by the Gleason score seems to be the best candidate to separate low- and high-risk groups, thus pointing further to the hope that more sophisticated biological markers in the tumor cells will be of great help in future clinical decision making. Acknowledgment The study was supported by the Swedish Cancer Society. References [1] Bill-Axelson A, Holmberg L, Ruutu M, et al. Radcial prostatectomy versus watchful waiting in early prostate cancer. N Engl J Med 2005;352:1977–84. [2] Jani AB, Hellman S. Early prostate cancer: clinical decision-making. Lancet 2003;361: 1045–53. [3] Holmberg L, Bill-Axelson A, Helgesen F, et al. A randomized trial comparing radical prostatectomy with watchful waiting in early prostate cancer. N Engl J Med 2002;347:781–9. [4] Mostofi FK, Sesterhenn I, Sobin LH. Histological typing of prostate tumors. In: International classification of tumors. WHO monograph series no. 22. Geneva: World Health Organization; 1980. [5] Johansson JE, Andren O, Andersson SO, et al. Natural history of early localized prostate cancer. JAMA 2004;291:2713–9. [6] Zietman AL, Thakral H, Wilson L, et al. Conservative management of prostate cancer in the prostate specific antigen era: the incidence and time course of subsequent therapy. J Urol 2001;166:1702–6. [7] Soloway M, Roach M. Prostate cancer progression after therapy of primary curative intent. A review of data from the prostate-specific antigen era. Cancer 2005;104:2310–22. [8] To ¨ rnblom M, Eriksson H, Franze´n S, et al. Lead time associated with screening for prostate cancer. Int J Cancer 2004;108:122–9. [9] Draisma G, Boer R, Otto SJ, et al. Lead times and overdetection due to prostate-specific antigen screening: estimates from the European Randomized Study of Screening for Prostate Cancer. J Natl Cancer Inst 2003;95:868–78. [10] Steineck G, Helgesen F, Adolfsson J, et al. Scandinavian Prostate Cancer Group Study number 4. Quality of life after radical prostatectomy or watchful waiting. N Engl J Med 2002;347:790–6.