International study into the use of intermittent hormone therapy in the treatment of carcinoma of the prostate: a meta-analysis of 1446 patients

Authors

  • Greg L. Shaw,

    1. St Bartholomew’s Hospital, and *Cancer Research UK, London, UK, †The Prostate Centre at VGH, Vancouver, Canada, and ‡Sir Charles Gairdner Hospital, and University of Western Australia, Perth, Australia
    Search for more papers by this author
  • Peter Wilson,

    1. St Bartholomew’s Hospital, and *Cancer Research UK, London, UK, †The Prostate Centre at VGH, Vancouver, Canada, and ‡Sir Charles Gairdner Hospital, and University of Western Australia, Perth, Australia
    Search for more papers by this author
  • Jack Cuzick,

    1. St Bartholomew’s Hospital, and *Cancer Research UK, London, UK, †The Prostate Centre at VGH, Vancouver, Canada, and ‡Sir Charles Gairdner Hospital, and University of Western Australia, Perth, Australia
    Search for more papers by this author
  • David M. Prowse,

    1. St Bartholomew’s Hospital, and *Cancer Research UK, London, UK, †The Prostate Centre at VGH, Vancouver, Canada, and ‡Sir Charles Gairdner Hospital, and University of Western Australia, Perth, Australia
    Search for more papers by this author
  • S. Larry Goldenberg,

    1. St Bartholomew’s Hospital, and *Cancer Research UK, London, UK, †The Prostate Centre at VGH, Vancouver, Canada, and ‡Sir Charles Gairdner Hospital, and University of Western Australia, Perth, Australia
    Search for more papers by this author
  • Nigel A. Spry,

    1. St Bartholomew’s Hospital, and *Cancer Research UK, London, UK, †The Prostate Centre at VGH, Vancouver, Canada, and ‡Sir Charles Gairdner Hospital, and University of Western Australia, Perth, Australia
    Search for more papers by this author
  • Tim Oliver

    1. St Bartholomew’s Hospital, and *Cancer Research UK, London, UK, †The Prostate Centre at VGH, Vancouver, Canada, and ‡Sir Charles Gairdner Hospital, and University of Western Australia, Perth, Australia
    Search for more papers by this author

Greg Shaw, Department of Urology, 2nd Floor Maple House, University College Hospital, London WC1E 5DB, UK.
e-mail: gregshaw@doctors.org.uk

Abstract

OBJECTIVE

To review pooled phase II data to identify features of different regimens of intermittent hormone therapy (IHT), developed to reduce the morbidity of treating metastatic prostate cancer, and which carries a theoretical advantage of delaying the onset of androgen-independent prostate cancer, (AIPC) that are associated with success, highlighting features which require exploration with prospective trials to establish the best strategies for using this treatment.

METHODS

Individual data were collated on 1446 patients with adequate information, from 10 phase II studies with >50 cases, identified through Pubmed.

RESULTS

Univariate and multivariate Cox proportional hazard models were developed to predict treatment success with a high degree of statistical success. The prostate-specific antigen (PSA) nadir, the PSA threshold to restart treatment, and medication type and duration, were important predictors of outcome.

CONCLUSIONS

The duration of biochemical remission after a period of HT is a durable early indicator of how rapidly AIPC and death will occur, and will make a useful endpoint in future trials to investigate the best ways to use IHT based on the important treatment cycling variables described above. Patients spent a mean of 39% of the time off treatment. The initial PSA level and PSA nadir allow the identification of patients with prostate cancer in whom it might be possible to avoid radical therapy.

Abbreviations
(I)HT

(intermittent) hormone therapy

AIPC

(androgen independent) prostate cancer

OS

overall survival

MAB

maximal androgen blockade

RP

radical prostatectomy

SEUG

Southern European Uro-Oncology group.

INTRODUCTION

It is now nearly 20 years since the first attempts were made to diminish the side-effects of continuous hormone therapy (HT) for prostate cancer by using pulsed endocrine treatments for long enough to stabilise disease activity. The side-effects of HT are well described and include anaemia, osteoporosis, impotence, cognitive functional effects, gynaecomastia, muscle atrophy, depression, dyslipidaemia and generalized lethargy [1–3]. The first series published by Klotz et al.[4] described the use of pulses of stilboestrol therapy for advanced prostate cancer with an improved side-effect profile has prompted further clinical and laboratory work to further investigate this type of treatment.

Work by Bruchovsky et al.[5] on Shionogi tumours in mice and Sato et al.[6] with LNCaP tumours in nude mice showed that by rendering the host mice castrate and then allowing the testosterone levels to return to pre-castrate levels for periods, androgen independence (AI) and death were delayed.

Stopping HT has the theoretical advantage of removing the selection pressure which, in a Darwinian fashion, selects the clones that have developed adequate molecular adaptations to allow them to have AI growth [7]. If there is a population of androgen-dependent clones then these will proliferate and repopulate the gland, and androgen dependence will resume.

During the past decade this preclinical evidence has prompted several phase II clinical studies of intermittent HT (IHT), although there has been considerable variation in approach. The consistent features of IHT are that HT is started, the PSA level is monitored and treatment is stopped when an adequate PSA nadir is reached. The PSA level is allowed to rise until it reaches a predetermined level, when HT is restarted or until clinical progression is evident. The patient has repeated cycles on and off treatment. The criteria for stopping and restarting treatment, as well as the medication types, vary among groups; some treat for a set duration rather than according to the PSA level. The ongoing randomized controlled trials have generated evidence that the use of IHT in patients with advanced or locally advanced disease is at least as safe as continuous HT [8,9].

The aims of the present meta-analysis are: (i) to develop models predicting the success in IHT and thereby identify features of IHT protocols to be the focus of future prospective trials; and (ii) to evaluate the use of time off treatment as a surrogate predictor of survival for use in future IHT trials.

METHODS

Pubmed was searched using the keywords ‘intermittent hormone/androgen ablation’, and the references from the papers found were checked and authors consulted as to other sources of data. Ten groups with data pertaining to the use of IHT (with >50 patients) were identified [10–19]. Table 1 shows the origin and basic characteristics of the patients in each study, and the characteristics of the IHT protocol. Individual patient data were collected from the authors and collated into a database of 1498 patients. Of these, 1446 had adequate data for inclusion in the present analysis. All patients were hormone-naive before starting IHT.

Table 1.  The contributing studies and IHT protocols
RefOriginNStage of diseaseType of treatmentPSA nadir, ng/mL for adequate responseRestart PSA, ng/mL% off at 2 years
  1. L, localized disease treated primarily with IHT; R, recurrent disease with no evidence of metastasis; A, advanced disease N+/M+ mono, monotherapy with LHRH/antiandrogen. RxT, radiotherapy.

[18]CA, USA 53L & AMAB<0.05>513
[19]MI, USA104L, R & AMAB/mono<4>1039
[11]Ottawa 86L, R & AMAB/mono<4>1020
[14]CA, USA 53L & RMAB/mono<4, NPT
<0.1 after RxT/RP
>10 or >50% baseline36
[12]Paris, France160L, R & AMAB/mono<1 NPT
<0.05 after RP
<4 after RxT
>10
>4 after RP
11
[13]Vancouver 101L, R & AMAB/mono<2>10
>4 after RP
17
[17]Perth, Australia239L, R & AMABVariable – all 9 months treatment>2031
[15]London, UK125L, R & AMAB/mono<4>2030
[16]Paris, France 411L, R & AMAB<4>2040
[10]Europe 114AMAB<20 or <20% initial>20 or
>1.5 × nadir
 6 (at 1 year)

Patient data were grouped into: group 1, localized disease treated primarily with IHT (N0/M0); group 2, PSA recurrence after failed curative attempts with radiotherapy, prostatectomy or both (N0/M0); group 3, metastatic disease treated with IHT (N+ or M+).

Data were analysed using univariate and multivariate Cox proportional hazard models, developed for each of the three groups. The variables studied were initial PSA level, type and duration of medication, PSA nadir, age, T stage, Gleason grade, previous treatment, the PSA threshold for restarting HT, and metastatic status; all durations are the length of time after starting HT. The variables were dichotomized for analysis of clinically relevant thresholds. A forward stepwise procedure was used to develop the multivariate models, and the Kaplan–Meier method was used for survival analysis.

The success of treatment was measured according to three factors: (i) the time spent in clinical remission and off treatment after the initial period of androgen deprivation; (ii) the time to AI prostate cancer (AIPC, defined as three successively increasing PSA levels or clinical progression despite HT); (iii) overall survival (OS).

RESULTS

Data were collected on 1446 patients (median age 71 years); 366 patients had confirmed nodal or metastatic disease at the time of starting IHT. Of the remaining 1080 with no evidence of metastasis, for 517 IHT was the primary treatment. The remaining 563 were treated for recurrent disease after failed radical prostatectomy (RP), radiotherapy or both. The median number of cycles was 2, the median time off treatment for all patients was 15.4 months, and the median follow-up for all patients was 39 months. Overall, patients spent a mean of 39% of time off treatment. The maximum follow-up was 197 months, during which 181 patients developed AIPC whilst 218 died.

Overall, 29% of patients with localized disease were off treatment 2 years after the initial period of hormone ablation was complete, 90% were alive at 5 years and 10% had become AI (Table 2). This compares to 33%, 86% and 17% in those with biochemical recurrence (group 2) and 16%, 68% and 41% in group 3. Table 3 shows a summary of the univariate analysis (see Appendix 1 and 2 for details). The results of the multivariate modelling are illustrated in Fig. 1.

Table 2.  Overall survival, time off treatment and time to developing AIPC for all patients by group
 Group
1 (localized disease, N0/M0)2 (biochemical recurrence, N0/M0)3 (metastatic disease N+ or M+)
Number of patients517563366
%:
OS at 5 years 90 86 68
Patients off treatment at 2 years 29 33 16
Patients with AIPC at 5 years 10 17 41
Table 3.  Summary of risk factors as predictors of outcome in univariate analysis by group
 Time off treatment, groupAndrogen resistance, groupOverall survival, group
123123123
  • *

    P < 0.05;

  • P < 0.01;

  • P < 0.001; RxT, radiotherapy.

T stage
 1/2
 3/41.48 1.85
Initial PSA level, ng/mL
 <10
 10–751.471.441.776.421.62*5.522.7
 >751.411.662.653.262.33
Grade
 2–7
 8–101.36
PSA nadir, ng/mL
 <1
 1–21.682.122.12
 >21.771.43
Type of medication
 MAB
 a/a2.14*2.59*3.78
 LHRH  4.41
Duration of treatment, months
 <41.86*
 4–84.44*2.83
 >86.05
Age, years
 <651.55
 ≥65
PSA restart threshold, ng/mL
 ≤15
 >151.33*1.65
Previous treatment
 RxT &/or RP2.2
Figure 1.

Kaplan–Meier survival analysis of risk factors: a , OS by group; b , duration of remission in patients with localised disease treated primarily with HT (group 1) by initial PSA level (ng/mL); c , OS by PSA threshold to restart treatment; d , the duration of remission in patients with localised disease treated primarily with HT (group 1) by duration of HT.

Table 3 shows the results of the univariate analysis for the range of risk factors studied as predictors of outcome. The tables in Appendix 1 and 2 show the details of the univariate analysis. The hazard ratios and P values are given for a particular risk factor when compared with the reference (Ref) value. Figure 1a shows the Kaplan–Meier survival analysis of patient groups as a predictor of OS. Figure 1b shows the power of PSA level before starting treatment as a predictor of the duration of remission. Of those with localized disease being treated primarily with HT, those with a PSA level of <10 ng/mL at the start of treatment have a considerably longer duration of remission on average. Of the subset of patients with a PSA level of <10 ng/mL, about half can expect to be off treatment for ≥ 2 years.

Figure 1c shows the effect on OS of the threshold to which the PSA level is allowed to rise before restarting treatment. Of the entire patient cohort those in whom treatment is restarted when the PSA level reaches 15 ng/mL survive longer than those in whom it is allowed to rise higher. This effect is highly statistically significant (P = 0.003). Figure 1d shows the effect of duration of treatment on the duration of remission; there was no difference in the duration of remission whether the patients (with localized disease) have a long treatment period (>8 months) or a short one (<4 months).

To assess whether the results from the univariate analysis hold when interaction with other variables is considered we used a multivariate analysis. The variables found to be predictors in the univariate analysis were entered into the multivariate model for analysis as candidate independent predictors of outcome. The resultant multivariate models are shown in Fig. 2, and are predictive of outcome at a highly significant level. The calculation for the modelling are shown in Appendix 3. The change in the chi-square value for the results of the multivariate analysis indicate the proportion of variability in outcome explained by the variable. A large change in the chi-square value indicates that a variable is important in predicting the particular outcome. The absence of a variable from a model indicates that it is not an important predictor of outcome for that group. The initial PSA level, PSA nadir achieved and type of medication are consistently important in predicting the outcome. The duration of treatment is only predictive in the patients with evidence of metastasis.

Figure 2.

Plots showing the relative contributions of individual factors in multivariate models predicting treatment outcomes.

To assess the restart threshold as a predictor of subsequent duration of remission, Table 4 shows a comparison in the rates of development of AIPC and OS for those patients in whom the PSA threshold for restarting treatment was <15 ng/mL, and those in who it was allowed to rise to >15 ng/mL, by univariate Cox modelling. Fig. 1c shows the Kaplan–Meier survival analysis by restart threshold for the entire cohort. There was a statistically significant difference in OS between those whose PSA was not allowed to rise to >15 ng/mL and those whose PSA was allowed to rise above this level before recommencing treatment. The multivariate models shown inFig. 2 show that this statistically significant difference is limited to group 3.

Table 4. 
The PSA restart threshold as a predictor of treatment success
PredictionPSA threshold for restart, ng/mL P
<15≥15
Off treatment at 1 year in subsequent cycle, %29270.83
Androgen independent at 5 years, %16280.018
OS at 5 years, %87790.003

Table 5 shows the association between the duration of remission and the other outcome measures, showing (the first category) that those patients with a period of remission of <2 years have a much greater risk of developing AIPC (hazard ratio 12.1) and dying (2.24), the ratio being highly statistically significant (P < 0.001).

Table 5. 
Time off treatment as an early indicator of treatment success
StatusHR (95% CI)
AIOS
Off treatment at 2 years
 No12.12 (5.93–24.74) 2.24 (1.56–3.23)
 YesRefRef
Chi-square, P97.77, < 0.00121.49, < 0.001
Patients who were neither off treatment for the duration of interest, 2 or 3 years, or restarted treatment excluded
Off treatment at 2 years
 No 9.5 (4.6–19.5) 2.2 (1.45–3.2)
 YesRefRef
Chi-square, P68.33, < 0.00115.76, < 0.001
Off treatment at 3 years
 No 6.82 (2.96–15.72) 2.93 (1.68–5.11)
 YesRefRef
Chi-square, P34.18, < 0.00117.73, < 0.001

One criticism of this analysis is that a bias would be introduced by including those who have been off treatment for <2 years, or those who have not yet restarted treatment. The second two categories in Table 5 shows the results of two further analyses where the patients who had not been off treatment for the required period or who had not yet restarted treatment were excluded. The analyses shown compare the hazard ratios for developing AIPC and dying in those (selected patients) who have had remission periods of <2 years and 3 years. The relationship that a shorter duration of remission is associated with greater risk of developing AIPC and death is robust throughout these three analyses.

DISCUSSION

In the present dataset, patients spent a mean of 39% of time off treatment. Multivariate models show the power of the initial PSA level and PSA nadir, and type of treatment and the PSA threshold for restarting treatment, in predicting outcome. Prospective trials to explore aspects of IHT that might be altered to improve outcome should focus on these variables. The duration of treatment was not an independent predictor of outcome in patients with no evidence of metastases. In those patients who rapidly achieve a good PSA nadir it is safe to curtail treatment to <4 months. In the presence of evidence of metastasis, treatment must be protracted to ≥ 8 months. Restarting treatment when the PSA level approaches 15 ng/mL is associated with improved survival in patients with metastases, indicating the need for a more aggressive treatment strategy in these patients. Maximum androgen blockade (MAB) or LHRH analogue should be the standard for patients treated with IHT.

From the time of the original observation on the use of hormonal manipulation as a treatment for prostate cancer [20] and the development of methods of medical castration, there has been controversy over when and for how long medical treatment should be administered. At present continuous HT is the standard of care only in patients with metastatic and poor-risk locally advanced disease, although there is increasing opinion that primary HT for selected patients with localized disease might be appropriate [21].

The results that have emerged from the ongoing phase three clinical trials [9,22] generated the impetus to conduct this meta-analysis. One of these trials, by the Southern European Uro-Oncology group (SEUG), which used an LHRH analogue and the antiandrogen cyproterone acetate to treat patients with metastatic and locally advanced disease, is the largest that has recruited to date. With 626 patients they showed equivalence between continuous and IHT for disease progression and prostate cancer-specific deaths with 8 years of follow-up. They used only 3 months of therapy before stopping treatment in the IHT arm [8]. Another smaller trial [9] of only 68 patients showed a statistically significant (P = 0.005) decrease in the rate of disease progression (7% vs 38.9%) in patients with advanced prostate cancer treated with IHT vs continuous HT, despite a short follow-up of only 30.8 months.

One long-standing criticism of IHT is that the time off treatment is entirely a result of the slowness of recovery of circulating testosterone. Studies show that 90% of patients who have been treated with 3 months of LHRH recover a normal testosterone level within 18 weeks [22]. In the SEUG trial, most sexually active men in the IHT arm recovered potency. These factors suggest that persistent low testosterone levels are not the explanation for the prolonged remission period.

The results of the present univariate and multivariate modelling are in agreement with the findings of the SEUG trial and confirm that stopping treatment in patients who have a good PSA response at 3 months is not deleterious to survival. We found that in those with radiographic evidence of metastasis, a treatment period of <8 months was associated with shorter survival. That monotherapy with antiandrogen alone was significantly less effective than MAB or LHRH alone in groups 1 and 3 suggests that for the time being MAB or LHRH alone should be the standard for IHT trials.

Using the present dataset, multivariate models were generated which predicted, very significantly, the time off treatment in all patient groups. From these models, we can conclude that the type of medication, initial PSA level and PSA nadir are consistent independent predictors of the duration of remission. Shorter periods of treatment are possible in those with no evidence of metastasis. These factors require assessment with prospective trials.

The level to which the PSA level is allowed to rise before treatment is restarted also predicted survival and the development of AIPC in all patients. Examination of the multivariate models shows that this effect is only important in patients with metastatic disease (group 3). This and the association between a short duration of treatment and poor outcome suggests that in these patients treatment needs to be more aggressive. The treatment period must be longer and medication restarted at a lower threshold than in patients who have no evidence of metastases.

A model to predict the development of AIPC and OS could only be generated for group 3 (those with metastatic disease) as the development of AIPC and death were too rare in the other groups. This indicates the long follow-up needed to show differences in AIPC/death in this patient group, and obviates the need for an early indicator of treatment success in these patients.

Table 5 shows the association between the time off treatment in the first cycle and time to develop AIPC (P < 0.001) and with time to death (not disease-specific. P < 0.001). Those with off periods of >2 or 3 years, or in the first cycle, survive longer and develop AIPC later than those with a shorter duration of off-treatment. This can be used as an early indicator of treatment success in trials, to show the effectiveness of different treatment protocols (with more or fewer patients staying off treatment for ≥ 2 or 3 years, depending on treatment efficacy) before showing a difference in the rate of AI or death, for which the duration of follow-up would be extensive.

The use of agents such as 5α-reductase inhibitors, which lower serum testosterone levels [23], and thereby lower PSA levels, to prolong the off-treatment period would necessitate the use of more traditional endpoints, AIPC and death.

A recent editorial described how prostate cancer is over-treated at present [24]. With the detection of prostate cancer more frequently of lower grade and at an earlier stage, with time the available treatment options must shift to reflect this. A short course of HT might identify those patients for whom the outcome would be good with this type of treatment, by identifying those with a good PSA response. In those with localized disease and who show a partial or poor response, radical treatment can then be used.

As the evidence showing the safety of IHT increases so do the number of trials and variety of approaches. The duration of treatment, the type of medication used, and the PSA thresholds for stopping and restarting treatment are largely based on personal experience and opinion. Only when treatment is optimized based on the results of these trials might there be the kind of effect observed in mice by Bruchovsky et al.[5], with a delayed onset of AI rather than merely equivalence to continuous treatment. The collated dataset lends itself to an analysis to evaluate which features of an IHT protocol render it most successful (within the bounds of those protocols already in use). The results will be useful in designing clinical trials examining the optimum use of IHT.

The limitations of the analysis are that the medication type was grouped despite acknowledging differences in different individual drugs and doses used. Testosterone levels were not available for enough patients for analysis.

In conclusion, the increasing evidence from randomized trials that IHT is safe is supported by the present meta-analysis. The results do not oppose the findings of Calais Da Silva et al.[8] that it might be possible to curtail the treatment duration to 3 months in all patients except those with metastatic disease. In patients with evidence of metastasis a more aggressive approach should be adopted, with a longer treatment period and earlier restart. In the prospective trials that are needed to investigate how best to treat with IHT, factors identified as having a major impact on the outcome are the initial PSA level, PSA nadir, and type and duration of medication. The type and duration of treatment, the PSA nadir sought and the PSA level prompting treatment restart are factors that can be defined in the treatment protocol. Our data suggest that the duration of remission is a useful early indicator of treatment success and could be used as an endpoint in these trials. Clearly using an endpoint such as duration of remission, which is determined by the PSA level, will only be possible when serum testosterone levels are known and controlled. Assessing quality of life and treatment side-effects are important endpoints for future studies [15].

ACKNOWLEDGEMENTS

This work was undertaken with support from the Orchid Cancer Appeal. We thank the European Organization for Research and Treatment of Cancer for permission to use data from EORTC trial 3095N4E for this research. The ISICAP group met in March 2005 and March 2006 to review preliminary analysis and drafts. We are grateful to the following people for providing patient data for the collated dataset and for their input into discussion, analysis and review of the manuscript: Prapotnich D, Dept of Urology, Université René Descartes, Faculté des Saints Pères, Paris, France; Forman J, Gerhenson Radiation Oncology Center of the Barbara Ann Karmanos Cancer Institute, Wayne State University, Detroit, Michigan, USA; Scholz MC and Strum S, The Prostate Cancer Research Institute, LA, California, USA; Albrecht W, Dept of Urology, Rudolfstiiftung, Juchgasse 25, Vienna, Austria. We are grateful to the following for providing patient data for the collated dataset: Small EJ, University of California, San Francisco, School of Medicine, San Francisco, CA, USA; Zerbib M, Department of Urology, CHU Cochin, Paris, France; Malone S, Department of Radiation Oncology, Ottawa Regional Cancer Centre, Canada.

Thanks to Prof. Urs Studer, Bern, for his careful review.

CONFLICT OF INTEREST

None declared. Source of funding: The Orchid Cancer Appeal.

Appendices

APPENDIX 1

Univariate analysis of risk factors for being off treatment at 2 years (group 1 and 2) or 1 year (group 3).

Table 6. 
Factorn, (%) off treatment; hazard ratio (95% CI), P
Group 1Group 2Group 3
  • *

    P values are for trend; RxT radiotherapy.

T stage
 1 or 2251, (29)212, (37) 54, (47)
 3 or 4187, (29); 1.04 (0.83–1.20), 0.8 181, (25); 1.48 (1.18–1.87), <0.001 72, (40), 1.02 (0.68–1.54), 0.9
 missing 37, (26) 113, (42)185, (28)
Initial PSA, ng/mL*
 <10 91, (41)175, (41) 23, (70)
 10-75295, (25); 1.47 (1.11–1.95), <0.001287, (27); 1.44 (1.14–1.92), 0.05163, (39); 1.77 (1.04–3.03), <0.001
 >75 60, (23); 1.41 (1.16–1.72) 21 (56), 0.95 (0.73–1.24) 116, (20); 1.66 (1.26–2.18)
 missing 71, (42) 23, (33)  9, (33)
Gleason grade
 ≤7165, (30)185, (38) 53, (51)
 >7237, (28); 1.26 (0.99–1.58), 0.05249, (29); 1.36 (1.08–1.71), 0.008133, (43); 1.25 (0.86–1.8), 0.02
 missing 67, (29) 72, (38)125, (17)
Type of medication antiandrogen 68, (11); 2.14 (1.23–3.24), 0.005 77, (10); 2.59 (1.83–3.72), <0.001 17, (0); 1.16 (0.88–1.36), 0.3
 LHRH 39, (31); 1.2 (0.96–1.45), 0.7 93, (38); 1.02 (0.93–1.10), 0.9 12, (39); 1.3 (0.77–1.82), 0.7
 MAB404, (32)389, (38)334, (15)
 missing  0  6, (42) 23, (26)
Duration on treatment, months*
 <4148, (35)125, (42) 75, (32); 1.35 (1–1.83), 0.6
 4–8102, (24); 1.26 (0.93–1.67), 0.03169, (28); 1.36 (1.03–1.81), 0.08 81, (27); 1.17 (1–1.82)
 >8219, (27); 1.06 (0.88–1.45) 211, (32); 1.1 (0.93–1.60) 151, (39)
 missing  0  1 (1/1)  1, (0)
PSA nadir, ng/mL*
 <1273, (33)395, (35)180, (41)
 1–2 67, (15); 1.68 (1.24–2.28), 0.003 32, (40); 0.85 (0.53–1.35), 0.049 30, (23); 1.36 (0.89–2.08), <0.001
 >2 29, (30); 1.45 (0.92–2.27) 18, (8); 3.63 (1.75–7.53) 49, (21); 1.43 (1.20–1.71)
 missing100, (25) 61, (25) 32, (31)
Age, years
 ≤65 70, (23); 1.16 (0.88–1.56), 0.3145, (25); 1.55 (1.25–1.94), <0.001102, (29); 1.19 (0.91–1.55), 0.2
 >65393, (30)353, (37)200, (38)
 missing  6, (42)  8, (43)  9, (1/9)
Previous treatment
 none281, (31); 2.2 (1.39–3.48), <0.001
 RP and/or RxT 30, (57)
 missing
PSA restart threshold, ng/mL
 <15190, (29); 1.03 (0.82–1.28), 0.8236, (30); 1.12 (0.91–1.38), 0.3 85, (32); 0.5
 ≥15294, (28)280, (35)271, (33); 1.1 (0.83–1.46)
 missing 33, (33) 42, (47) 10, (27)

APPENDIX 2

Significance of factors predicting time off treatment in the different clinical groups, by univariate analysis

Table 7. 
FactorHazard ratio (chi-square), Group n
Time off treatmentAIOS
123123123
  • *

    P < 0.05;

  • P < 0.01;

  • P < 0.001. R, reference; N/A, not applicable; –, not significant; RxT, radiotherapy.

T stage
 1–2RR
 3–41.48 (11.3)1.85 (3.31)
Initial PSA level, ng/mL
 <10RRRRRRRR
 10–751.47 (7.8)1.44 (9.9)1.77 (5.06)6.42 (5.55)1.62 (2.4)*5.52 (5.09)2.70 (4.21)
 >751.41 (11.2)1.66 (16.1)2.65 (3.96)3.26 (5.23)2.33 (9.42)
Gleason grade
 ≤7R
 >71.36 (7.15)
Type of medication
 MABRRR2.28 (6.12)*
 LHRH alone4.41 (5.82)*4.28 (13.7)
 antiandrogen2.14 (19.1)2.59 (39.3)3.78 (4.84)*R
Duration of treatment, months
 <4R1.86 (2.89)*
 4–84.44 (2.4)*2.83 (13.5)
 >86.05 (5.1)R
PSA nadir, ng/mL
 <1RRRRR
 1–21.68 (10.9)2.12 (6.98)2.12 (6.98)
 >21.77 (10.7)1.43 (13.9)
Age, years
 <651.55 (14.6)
 ≥65R
Previous treatment
 NoneN/AN/ARN/AN/AN/AN/A
 RP +/or RxT  2.2 (14.0)    
PSA restart threshold, ng/mL
 <15RR
 ≥151.33 (4.11)*1.65 (4.08)*

APPENDIX 3

Multivariate model for each group predicting the duration of remission (time off treatment) based on initial PSA level (IP), PSA nadir <1 (PN) reached, type of HT (MAB vs monotherapy, m), tumour Gleason grade (G), T stage (T), age (A), previous treatment (PrevRx), restart PSA threshold (RT).

ModelNChi-squareChange in chi-square
Group 1; Time off treatment
IP446 9.67 9.67
IP + IPm46412.31 
IP + IPm + MAB46429.9217.61
IP + IPm + MAB + MABm36937.5 
IP + IPm + MAB + MABm + PN36946.95 9.45
IP + IPm + MAB + MABm + PN + PNm40238.84 
IP + IPm + MAB + MABm + PN + PNm + G40246.05 7.21
Group 2; Time off treatment
MAB50113.9813.98
MAB + MABm 49714.58 
MAB + MABm + A49729.0614.48
MAB + MABm + A + Am48225.09 
MAB + MABm + A + Am + IP48231.46 6.37
Group 3; Time off treatment
IP30223.8623.86
IP + IPm31123.89 
IP + IPm + PrevRx31135.6 11.71
IP + IPm + PrevRx + PN25939.05 4.55
Group 3; AI
IP35524.4824.48
IP + IPm34627.23 
IP + IPm + RT34630.07 5.59
Group 3; OS
ON32913.7213.72
ON + ONm28010.39 
ON + ONm + PN28017.09 3.37
ON + ONm + PN + PNm27717.04 
ON + ONm + PN + PNm + IP27723.27 6.18
ON + ONm + PN + PNm + IP + IPm27524.86 
ON + ONm + PN + PNm + IP + IPm + MAB27530.42 7.15
ON + ONm + PN + PNm + IP + IPm + MAB + MABm26728.16 
ON + ONm + PN + PNm + IP + IPm + MAB + MABm + RT26737.03 6.61

Ancillary