This systematic review and meta-analysis shows that many treatments for CP/CPPS are largely ineffective. Our results demonstrate several critical issues underlying the limited success of clinical trials in CP/CPPS, including the roles of placebo and treatment duration.
Single trials of mepartricin,
and combination DIT
showed at least a six-point overall reduction in total NIH-CPSI score. Several methodological limitations preclude the generalizability of those findings, namely: inadequate blinding and the subsequent allocation bias (De Rose et al.
), imperfect allocation concealment (Kabay et al
and Tuğcu et al.
), inappropriate randomization, inadequate withdrawal reporting, and lack of detailed reporting of adverse effects (Kabay et al
). Most concerning in the Kabay et al
study was the fact that while the treatment group received the maximum tolerable electrical stimulation, the “sham group” did not meet the definition of a sham since they did not receive any stimulation and were not blinded, prompting concerns that participants were able to differentiate between treatment and “sham.” This may be reflected in both in the efficacy of the treatment group as well as nearly absent “placebo effect” in the control group.
The alpha blocker trials included in the present meta-analysis demonstrated a high degree of heterogeneity (I2
76.3% for total NIH-CPSI and I2
61.6% for the NIH-CPSI Pain Subscore). We explored several sources of this heterogeneity which might have introduced bias in the design, execution and subsequent interpretation of those RCTs.
First, we examined the variability in the rigorousness of application of the eligibility criteria used for patient selection at randomization. We distinguished three levels of sensitivity and rigor of definition: CPCRN criteria, NIH criteria and “unclear”. For example, some studies explicitly stated that they employed the CPCRN criteria.
referred to the JAMA 1999 Krieger criteria
as “the NIH criteria” which, although similar to the CPCRN criteria, do not specify conditions exclusionary to the diagnosis of CP/CPPS and are, therefore, less rigorous. Several studies did not clearly describe the criteria used for defining the patient population.
Studies that failed to use CPCRN or NIH criteria reported greater effectiveness of treatment.
Secondly, we examined the importance of patient age. Generally, it has been accepted that CP/CPPS patients younger than 50 might respond differently to alpha-blocker treatment than those older than 50, mostly as a result of biological variables, such as the increased prevalence of BPH in the older patient population. Our analysis found that older patients had less response to CP/CPPS symptoms than younger patients, a paradoxical finding. This finding should be interpreted with caution for several reasons. First, this analysis is based on average age among all participants in each specific study; fully elucidating age effects on treatment response would require patient-level data. Secondly, the average age of participants in each study was constrained and relatively young, ranging from 29 to 50 years in age. Benign prostatic hyperplasia is not a significant problem in men of this age, reducing potential benefit from alpha-blockers on BPH. Given these limitations, at least four possible explanations can be evoked for this finding: 1) Longer symptom duration in older patients prior to treatment initiation; or 2) Older patients preferentially enrolled in trials of shorter duration; or 3) Older patients that were already treatment refractory at study initiation; or 4) Publication bias. While our analysis did not reveal any preferential enrollment based on age (thus, arguing against the first explanation) or treatment refractoriness at baseline (thus, arguing against the second one as well), our analysis might have been limited by the lack of detail in terms of age or treatment-based enrollment status. We did not detect any age-based publication bias either. Thus, the most plausible explanation appears to relate to the fact that longer duration of symptoms in older patients might explain the paradoxical less favorable response to treatment. Our findings are in line with the Turner study in which men with a first lifetime episode of prostatitis/pelvic pain syndrome had better outcomes compared with men with a recurrent episode.
The best and only available review on the natural history of CP/CPPS comes from Kusek and Nyberg from the NIH/NIDDK.
They observed that “among nearly 300 men with longstanding CP/CPPS (mean 6.8 yr since diagnosis) recruited from tertiary care centers and followed for 2 yr in the CPCRN study, 45% reported that they were markedly or moderately improved on the Global Response Assessment. Importantly, there was no evidence of clinically significant progression of symptoms. These investigators, however, failed to identify baseline demographic or clinical studies which predicted improvement.”
. In addition, the same authors reference the Nickel et al.
reporting of all men in the community identified to have prostatitis-like symptoms, approximately one-third did not report these symptoms one year later. The preceding studies, as well as our finding of age-related paradoxical treatment response, related to the phenomenon of “prostatitis burning itself out” and a decrease in symptom severity with longer duration, thus causing a greater degree of regression to the mean in older patients. Future epidemiologic studies should provide further insight into mechanisms underlying this intriguing finding.
In addition, it has been hypothesized that alpha-blocker-naïve patients might respond differently (better) to alpha-blockade than those with previous alpha-blocker exposure. Some trials explicitly stated that “patients were excluded if they had ever previously taken alpha-blockers”
, while others stated that patients were excluded only if they had taken an alpha-blocker within a pre-specified period of time before study enrollment.
We believe this distinction might be important and might give rise to substantial heterogeneity in effect size since the alpha-blocker-naïve population might be different from the “alpha-blocker washout” population. We found no impact of pre-treatment with alpha-blockers, either for all studies or for those studies focusing on alpha-blocker treatment.
Other important variables included duration of study and a number of quality markers including lack of adequate sequence generation and allocation concealment. Weaker studies reported greater effectiveness with treatment. This has been seen previously in systematic reviews and including weaker trials in analyses may overestimate potential benefit. In our study, we found little benefit.
Our analysis raises the issue of placebo effect in CP/CPPS clinical trials. The improvement for the pool of all placebo groups is significant for CP/CPPS symptoms overall and for all three NIH-CPSI domains. For pain, such findings are consistent with a larger body of literature that has shown contextual elements of treatment to have the most powerful effects specifically in analgesia. Without a no-treatment comparison group, however, the placebo groups in these studies also capture the improvement of symptoms due to the natural history of an unstable course of illness. The lack of change in improvement over time for the placebo groups, however, weakens natural history as a sole explanation for these results. Furthermore, previous studies of CP/CPPS have suggested placebo effect can be significant, at least over the short term, for up to three months. The placebo-controlled studies in the present meta-analysis, with an average length of 13.4 weeks, are not of sufficient length to conclude whether placebo effect might wane in the long term.
While double-blind RCTs have long been the “gold standard” for limiting bias in clinical medicine, they are not without their own limitations and biases.
Their underlying assumption is that by blinding both provider and patient, the placebo effect arising should be identical in both groups and, therefore, impact findings from both groups equally. However, this assumption has been challenged. One study in particular found a clinically significant difference between albuterol treatment and three control treatments (two placebos and no treatment) when patients were evaluated using a “hard” outcome variable, FEV1.
Surprisingly, the clinical significance between the interventions evaporated when looking at a “soft” outcome variable, a subjective patient response questionnaire. This suggests that for “soft” outcome variables the placebo effect in the control group is greater than the placebo effect in the treatment group, thereby undermining an underlying assumption of double-blind RCTs.
All studies included in this meta-analysis use a “soft” primary outcome variable, the NIH-CPSI score. If a similar phenomenon occurred in these studies and the placebo effect is more substantial in control than treatment groups, then the treatment effect could be blunted.
Given these considerations, it appears that there was a significant placebo effect captured in the studies included in this analysis. This itself is noteworthy because it suggests that contextual elements of care play a measurable role in patient improvement for a condition not easily treated. Patients with a chronic pain condition have often tried many failed treatments, a history of which could in turn produce negative expectations for new therapies. Therefore, why a placebo effect exists for CP/CPPS warrants further investigation because expectancy of pain relief is one primary mechanism of placebo analgesia. Qualitative research with participants in CP/CPPS trials may be the most effective manner to gain such an understanding.
There are several important limitations to our review. First, for nearly all modalities demonstrating efficacy, there were only small, single-center trials. Particularly for a syndrome with such a powerful placebo effect, replication and confirmation of efficacy needs to be done before any modality can be conclusively stated to be helpful. Second, there was a wide range of study quality. Several trials had questionable placebo groups and inadequate blinding. This makes interpretation of the results difficult and may lead to erroneous conclusions. This reiterates the importance of replication and confirmation. Third, there may be specific patient subsets, which might respond to specific medications. The heterogeneity of CPPS makes identifying this subgroup difficult. The “therapeutic efficacy” of certain medications might be “diluted” because responders may be overwhelmed by non-responders.
Lack of patient-level data made it difficult for us to fully explore such potential subgroups. Fourth, most of the studies were relatively short, averaging only 13.4 weeks in duration. While our analysis suggested continued improvement over time, the longest studies were 32 weeks; it is impossible to speculate what might happen with longer studies. Patients could continue to improve, they could plateau or they could worsen. Longer studies are needed. Finally, we were not able to analyze based on specific, important patient characteristics such as disease duration or inclusion of treatment refractory versus treatment naïve patients. This would require patient-level data or the analysis to be by the authors in the individual manuscripts. It would be nice for the authors to make an assessment of the effect of disease duration in patients (<2 yrs versus >2 yrs) and inclusion of treatment refractory versus treatment naive participants on the effectiveness of the various therapies. It is possible that patients with longer duration or previously refractory treatment would be less responsive than those with shorter disease duration or treatment naïve.
Our study also shows the potential importance of treatment duration. In particular, the pooled data for alpha-blockers showed increased efficacy over time in addition to statistically significant improvements for overall NIH-CPSI and all three subdomains. Given the relatively short average study length and the 4.5 point average reduction in total NIH-CPSI for alpha-blockers, longer studies of alpha-blockers might result in clinically significant reductions in NIH-CPSI total score if the trend continues over time. For the pooled treatment data, the NIH-CPSI score decreased an average of 0.23 points each week, suggesting that 26 weeks of treatment would be required to see a six point reduction in the total NIH-CPSI score. Our study has significant difference from a recently published network meta-analysis (Table S9
The mixed results of the studies in this meta-analysis highlight the heterogeneity of CP/CPPS and our current lack of understanding of the etiology of the disease. Furthermore, our results highlight the significant limitations of previous trials, the existence of a potentially important placebo effect and the need for further quantitative and qualitative research in CP/CPPS. The observed variability in response to therapy could suggest that CP/CPPS is actually comprised of a number of separate disease entities with discrete causes that require different treatments. Our current understanding of CP/CPPS is not complete enough to allow us to employ appropriate interventions for all patients and it is important to continue to conduct research to improve our understanding of the mechanism and treatment of the disease.