Impact of early vs. delayed initiation of dutasteride/tamsulosin combination therapy on the risk of acute urinary retention or BPH-related surgery in LUTS/BPH patients with moderate-to-severe symptoms at risk of disease progression

Purpose To evaluate the effect of delayed start of combination therapy (CT) with dutasteride 0.5 mg and tamsulosin 0.4 mg on the risk of acute urinary retention or benign prostatic hyperplasia (BPH)-related surgery (AUR/S) in patients with moderate-to-severe lower urinary tract symptoms (LUTS) at risk of disease progression. Methods Using a time-to-event model based on pooled data from 10,238 patients from Phase III/IV dutasteride trials, clinical trial simulations (CTS) were performed to assess the risk of AUR/S up to 48 months in moderate-to-severe LUTS/BPH patients following immediate and delayed start of CT for those not responding to tamsulosin monotherapy. Simulation scenarios (1300 subjects/arm) were investigated, including immediate start (reference) and alternative delayed start (six scenarios 1–24 months). AUR/S incidence was described by Kaplan–Meier survival curves and analysed using log-rank test. The cumulative incidence of events as well as the relative and attributable risks were summarised stratified by treatment. Results Survival curves for patients starting CT at month 1 and 3 did not differ from those who initiated CT immediately. By contrast, significant differences (p < 0.001) were observed when switch to CT occurs ≥ 6 months from the initial treatment. At month 48, AUR/S incidence was 4.6% vs 9.5%, 11.0% and 11.3% in patients receiving immediate CT vs. switchers after 6, 12 and 24 months, respectively. Conclusions Start of CT before month 6 appears to significantly reduce the risk of AUR/S compared with delayed start by ≥ 6 months. This has implications for the treatment algorithm for men with LUTS/BPH at risk of disease progression. Electronic supplementary material The online version of this article (10.1007/s00345-020-03517-0) contains supplementary material, which is available to authorized users.

* An overview of the definition of BPH-related surgery is summarised in Table S2. AUR Table S5 Protocol design characteristics used for the simulation* of AUR/S and individual IPSS trajectories after immediate and delayed start of treatment with tamsulosindutasteride combination therapy:

Protocol characteristics
Endpoints: Endpoints evaluated were: 1) time to AUR or BPH-related surgical intervention; 2) incidence of AUR or BPH related surgical intervention (at different time points relative to the start of treatment); 3) increase in the time to AUR or BPH-related surgery due to combination therapy; 4) relative risk of AUR or BPH-related surgery in patients immediately treated with combination therapy, as compared to alternative treatments, namely tamsulosin monotherapy and delayed onset of combination therapy; 5) attributable risk due to the delayed onset of treatment with tamsulosin and dutasteride combination therapy (CT).

Simulation scenarios
Simulation scenarios were used to evaluate the long-term effect of different intervention on time to AUR/S in BPH patients with moderate or severe symptoms at baseline. To ensure simulations reflected the original clinical trial population, baseline characteristics from the available pooled population (Table S2) were resampled and used as input in each of the simulation scenarios. In addition, re-sampling of patients into each clinical trial scenario was performed taking into account the observed covariate distributions observed in the available clinical studies. Given the large sample size used across the different scenarios, along with the prior evidence of the predictive performance of the model, results from 10 trial replicates were assumed to be sufficiently accurate and precise to assess eventual differences between scenarios. Confidence intervals from trial replicates based on resampling from the same patient pool were deemed appropriate to account for model parameter uncertainty.
In an attempt to mimic clinical practice, a parallel study design was used, i.e., each patient was randomised to one of the treatment arms at the beginning of the trial. The transition from tamsulosin to CT in predefined study arms was based on LUTS symptom deterioration. Only non-responders to tamsulosin were assigned to CT, as assessed by the relative changes in IPSS scores at the predefined visit. Treatment response (LUTS improvement) was defined as a drop in IPSS score of ≥25% from baseline. Consequently, a non-responder is a subject who shows a change in IPSS score of < 25% relative to baseline. Except for baseline values, which were resampled from the original patient pool, individual IPSS values were derived by simulation using the drug-disease model developed previously [1]. In addition to the incidence of events, Kaplan-Meier curves, relative risk and the attributable risk of AUR and BPH-related surgery were also evaluated.

Baseline characteristics:
The patient population randomised into each simulation scenario was based on the inclusion and exclusion criteria used in the CombAT and CONDUCT studies [2,3]. Inclusion criteria comprised the following requirements at baseline: men ≥50 years of age, IPSS ≥8 points, prostate volume ≥30 mL, total serum PSA ≥1.5 ng/mL and <10 ng/mL, Q max >5 ml/s and ≤15 mL/s and a minimum voided volume ≥125 mL. Exclusion criteria included history or evidence of prostate cancer, previous prostatic surgery, history of AUR within 3 months prior to study entry, 5ARI use within 6 months (or dutasteride within 12 months) prior to entry or use of an α-blocker or phytotherapy for BPH within 2 weeks prior to entry.

Sample size estimation
Sample size calculations were based on the estimates of AUR/S incidence at month 48, as reported by Roehrborn and colleagues [3]. Post-hoc sample size was computed under the assumption of an incidence of 4.2% for CT. A conservative sample size of 1300 subjects per arm (total: 9100 subjects) was chosen to detect a statistically significant and clinically relevant difference of 3% in TTE rate between treatment arms at month 48. Considering the aforementioned effect size, 0.05 level of significance and 90% statistical power, the minimum sample size required for the analysis was 1260 subjects per arm.

Study visits
Simulation scenarios were based on typical clinical visits and included follow-up over a period of 48 months. This period was used to avoid empirical extrapolation beyond the observation window considered for model development and validation, for which there is also supporting clinical trial data. Visits including IPSS measurements were at study entry (baseline) and at 1, 2, 3, 4, 5, 6, 12, 18, 24, 30, 36, 42 and 48 months after the start of treatment. Screening measurements were not included or considered for simulation purposes. To minimise censoring and reduce eventual bias when tabulating the results, additional sampling times were simulated at bi-weekly time intervals for the evaluation of occurrence of AUR/S events.

Treatment arms
For this analysis, patients were assumed to have undergone a washout period prior to the start of the intervention. In addition, the assigned treatments (0.5 mg dutasteride and 0.4 mg tamsulosin) were initiated as CT or tamsulosin monotherapy under the assumption of varying intervals or delay after BPH diagnosis, as in the original patient population. One reference and six alternative treatment scenarios were evaluated, namely: Scenario 1 (reference): CT over 48 months; Scenario 2: tamsulosin over 48 months; Scenario 3: tamsulosin until visit 2 (1-month delay) followed by CT up to month 48; Scenario 4: tamsulosin until visit 4 (3-month delay) followed by CT up to month 48; Scenario 5: tamsulosin until visit 7 (6-month delay) followed by CT up to month 48; Scenario 6: tamsulosin until visit 8 (12-month delay) followed by CT up to month 48; Scenario 7: tamsulosin until visit 10 (24-month delay) followed by CT up to month 48.
Only non-responders to tamsulosin switched to CT at the selected visits.

Statistical methods:
Simulated events (AUR/S) were described by Kaplan-Meier survival curves and analysed using a log-rank test. Survival curves are a standard way to represent the occurrence of an event over a predefined observation window [4]. In the context of this analysis, survival refers to the proportion of patients who have not had an AUR/S. Median estimates and 90% confidence intervals are presented in tabular format for results from trial replicates (n=10). On the other hand, to ensure direct comparison with data arising from a prospective study protocol, graphical summaries are described as those obtained in a single trial, i.e., only point estimates are shown without confidence intervals.
The primary objective of the statistical analysis was to compare the difference in the incidence of AUR/S at month 48 between patients in the reference treatment arm (i.e. those who initiate treatment immediately with tamsulosin and dutasteride CT) and those whose treatment with tamsulosin and dutasteride CT is delayed. Assessment of the superiority of immediate onset of treatment with tamsulosin and dutasteride CT was based on a two-sided test with significance level α=0.05. As transition from tamsulosin to CT was implemented by design, i.e., with transition defined at prespecified times for each treatment arm, no additional statistical methods for adjustment of estimated treatment effect were used to correct for potential bias in estimates due to patients switching from their allocated treatment [5,6]. In addition to the incidence of AUR/S at month 48, attributable risk, relative risk and time to reach comparable incidences at 48 months following different interventions were also calculated.
Given that the endpoints of the clinical trial simulations were only considered at the end of study, the hazard ratio was not estimated when summarising the results. Moreover, since the rate of events is low and the incidence of events corresponding to a survival rate of 50% of the subjects is never reached during the time frame considered for the simulations, the median time to event could not be calculated with enough precision and was therefore excluded from the analysis. Instead, the time at which the incidence of AUR/S after delayed initiation of CT reaches that of the reference group at 48 months was estimated to assess the implications of treatment with disease-modifying properties on the progression of BPH.

Assumptions:
and limitations • Parameter estimates obtained from the pooled patient database (N=10,234 had dosing records) were assumed to be sufficiently precise to replicate the performance of the treatment in a wider population, as observed in clinical practice. However, we recognise that inclusion/exclusion criteria may not fully reflect the LUTS/BPH population eligible for treatment with α-blockers and 5-ARI in clinical practice.
• Interindividual variability in pharmacokinetics was assumed to have a minor impact on treatment response, as the currently approved dose levels of tamsulosin-dutasteride yield nearly maximum pharmacological effect.
• All CTS scenarios were implemented under the assumption of perfect adherence to treatment. In real-life conditions, different adherence patterns may occur, depending on symptom severity and/or comorbidities, which may significantly alter the predicted differences across CTS scenarios • As drop-out in real clinical trials appears to be non-informative (i.e. at random), treatment scenarios were implemented without dropout.
• The cumulative incidence was not calculated beyond 48 months to ensure that simulation results could be supported by existing clinical data, i.e., the time span used in the analysis matches the duration of the longest clinical trial included in the development of the model.
• The use of 10 trial replicates may seem limited, but given the sample size in each treatment arm (n=1300), this was deemed adequate to obtain reliable estimates of the 90%-confidence intervals of the survival function.
• The assessment of treatment response at each visit using predicted IPSS for each patient was based on the longitudinal model describing individual IPSS trajectories. Whilst this does not represent a limitation, the predicted response does not include residual random variability, which may be relatively large in real life.
5g-ARI, 5α-reductase inhibitor; AUR, acute urinary retention; CTS, clinical trial simulations; IPSS, International Prostate Symptom Score; LUTS/BPH, lower urinary tract symptoms due to benign prostatic hyperplasia; PSA, prostate-specific antigen; Q max , maximum urine flow rate.  Table S3. AUR/S events were simulated for each scenario along with IPSS response over a period of 48 months. All required data manipulation, including graphical and statistical summaries were performed in R (version 3.1.1).

Table S6
Demographics and baseline characteristics of the patient population used for the evaluation of the effect of early vs. delayed onset of treatment with combination therapy using clinical trial simulations.
Total number of patients at each visit includes those who switch to CT. Adherence to treatment and was considered constant throughout the duration of the study. Dropout was considered zero for the purposes of the analysis.

Combination therapy (CT) Tamsulosin
Tamsulosin nonresponders to CT at 1 month Tamsulosin