Meta-analyses have been used to compare different treatment options. However, this technique allows scrutiny only of design factors that have been implemented, without necessarily correcting for the effect of confounding factors which cannot be easily excluded. Moreover, they often focus on mean parameter estimates, yielding results that ignore underlying covariates that may modify the treatment effect. By contrast, the application of longitudinal modelling and CTS at individual patient level allows investigation of a range of design characteristics on the power to detect treatment effects, without confounding or practical restrictions, prior to exposing patients to an experimental intervention [
16,
17]. CTS can be performed not only to evaluate scenarios that have not been previously investigated in clinical trials, but also to explore hypothetical scenarios which cannot be implemented in real-life conditions. Indeed, the implementation of a prospective, controlled study in which combination therapy is delayed may be ethically questionable, especially when guidelines recommend it in patients considered at risk for progression of BPH [
4]. Here we have shown how this methodology can be used to explore design factors, such as delayed start of treatment, whilst disentangling it from other factors and interactions. Our analysis also provided an opportunity to assess the effect of disease progression, baseline covariates, and drug treatment on individual IPSS trajectories.
Effect of disease progression, baseline covariate factors and drug treatment on individual IPSS trajectories
Notwithstanding the body of evidence regarding the benefits of tamsulosin–dutasteride combination therapy, including greater and more durable improvement than with either monotherapy [
14,
18,
19], little attention has been given to the impact of variable rates of disease progression on treatment response or deterioration of symptoms, as measured by IPSS [
10,
20,
21]. There are currently no reliable biomarkers that allow identification and prediction of a specific clinical phenotype for disease progression in individual patients, although serum PSA has been explored in this capacity [
8,
9]. This is further compounded by limited understanding of the effects of specific comorbidities or other covariates on overall treatment response [
22]. Our analysis suggests these limitations may partly be overcome by further characterisation of individual IPSS trajectories.
The introduction of IPSS as a tool for clinical practice and in research protocols was originally based on data from relatively short-term validation steps [
23]. Among the available reports on the natural history of LUTS, long-term longitudinal follow-up studies have been restricted to changes in IPSS relative to baseline, making it difficult to distinguish the impact of multiple interacting factors on the deterioration of symptoms. Exploratory investigations have suggested that IPSS does not correlate with laboratory measures of urinary function at baseline (e.g., PV and PSA), but limitations in the methodology employed indicate that this should not be taken as conclusive [
24,
25]. In fact, the lack of such correlations has been often assigned to study design, inclusion and exclusion criteria, and sample size. Nevertheless, baseline IPSS has been shown to predict the risk of progression in several prior clinical trials [
6‐
8].
These apparently paradoxical findings may reflect the empirical nature of past research protocols, which assess or infer disease progression directly from the observed increase in IPSS over time, rather than based on a specific parameter that captures the underlying processes [
26‐
28]. In this context, the initial simulations show the implications of the interaction between baseline covariates (e.g. IPSS) and varying rates of disease progression on individual trajectories (see Supplemental Materials). Although the possibility that patients with LUTS/BPH may show intrinsic differences in response to α-blockers and 5ARI cannot be excluded, it is clear from figures S1, S2 and S3 that the identification of inter-individual variation in progression rates only from baseline characteristics may be challenging. Further exploration is required to identify whether baseline markers may explain some of this heterogeneity in symptom deterioration and consequently in treatment response.
Clinical trial simulations: impact of immediate versus delayed start of combination therapy
While the use of combination therapy in LUTS/BPH patients with moderate and severe symptoms at risk of disease progression is endorsed by international guidelines, this analysis was performed taking into account prior evidence of the favourable benefit–risk profile of dutasteride–tamsulosin combination [
29]. An early start of treatment with combination therapy should not alter the overall safety profile. However, the current results show that symptomatic treatment does not stop disease progression, and potentially reduces the benefits associated with the 5-ARI disease-modifying properties, if treatment is delayed for more than 6 months.
To address the key research question of this investigation, we have focused on the results of a parallel-study design, which reflects a typical clinical trial setting in which patients are randomised to different treatment arms, even though it does not fully capture the implications of delayed start of combination therapy for every single patient. In this regard, our results do not exclude the contribution of between-patient variability on predicted treatment response. It is worth mentioning that baseline characteristics among tamsulosin non-responders do not differ significantly from patients on combination therapy raising questions as to whether baseline demographics and clinical features may be sufficiently discriminative in determining treatment response. Yet, early administration of tamsulosin–dutasteride combination therapy (i.e. < 6-month delay) not only results in a significantly greater responder rate at month 48 in men at risk of progression, but also leads to an increase in the fraction of patients transitioning to lower IPSS severity levels. More strikingly, early use of combination therapy results in a larger proportion of patients having greater clinical improvement, i.e. changes in IPSS relative to baseline (Table
1, panels C and D). Early start of combination treatment allows ~ 10% more patients to benefit from symptomatic improvement.
The clinical relevance of differences in IPSS may be questioned without careful understanding of the approach used here. First, the definition of response based on a ≥ 25% reduction in IPSS relative to baseline allows for a standardised measure across the entire population, irrespective of symptom severity at baseline, provided that the response strongly influences the relationship between the perception of improvement and the score improvement. This threshold of ≥ 25% reduction in IPSS has been used as a study endpoint in various clinical trials included in this CTS [
14,
15]. Second, CTS were performed without residual noise to ensure accurate estimates of the effect of the different interventions. Residual noise reduces the sensitivity to detect a true treatment response.
Even though a few assumptions are required to assess treatment response when designing a standard clinical trial protocol or a virtual one, as in our case, our results highlight that immediate exposure to tamsulosin–dutasteride combination therapy can have a long-lasting impact on the individual trajectories of a small, but clinically relevant fraction of patients (i.e. those who have a faster disease progression rate but cannot be identified at start of treatment). Considering the chronic nature of the disease, the effect of disease-modifying properties cannot be compensated by symptomatic interventions over the longer term.
Undeniably, there are limitations in the work performed. First, it should be recognised that the population used for the development of the longitudinal model represents a subgroup of the overall BPH patient population, and as such encompasses only subjects with moderate or severe symptoms at risk of progression. In addition, we have assumed that the placebo effect starts at the beginning of the trial and not when the therapy is switched at later time points. As the longitudinal model used for the simulations showed some bias in terms of the predicted trajectories severe patients with very high IPSS scores (i.e., upper 2.5th percentile), there is a chance of misclassification of non-responders. Last, we have not evaluated IPSS profiles 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.