Economic evaluation of treatments for patients with localized prostate cancer in Europe: a systematic review

Background Our objective was to assess the efficiency of treatments in patients with localized prostate cancer, by synthesizing available evidence from European economic evaluations through systematic review. Methods Articles published 2000–2015 were searched in MEDLINE, EMBASE and NHS EED (Prospero protocol CRD42015022063). Two authors independently selected studies for inclusion and extracted the data. A third reviewer resolved discrepancies. We included European economic evaluations or cost comparison studies, of any modality of surgery or radiotherapy treatments, regardless the comparator/s. Drummond’s Checklist was used for quality assessment. Results After reviewing 8,789 titles, 13 European eligible studies were included: eight cost-utility, two cost-effectiveness, one cost-minimization, and two cost-comparison analyses. Of them, five compared interventions with expectant management, four contrasted robotic with non robotic-assisted surgery, three assessed new modalities of radiotherapy, and three compared radical prostatectomy with brachytherapy. All but two studies scored ≥8 in the quality checklist. Considering scenario and comparator, three interventions were qualified as dominant strategies (active surveillance, robotic-assisted surgery and IMRT), and six were cost-effective (radical prostatectomy, robotic-assisted surgery, IMRT, proton therapy, brachytherapy, and 3DCRT). However, QALY gains in most of them were small. For interventions considered as dominant strategies, QALY gain was 0.013 for active surveillance over radical prostatectomy; and 0.007 for robotic-assisted over non-robotic techniques. The highest QALY gains were 0.57–0.86 for radical prostatectomy vs watchful waiting, and 0.72 for brachytherapy vs conventional radiotherapy. Conclusions Currently, relevant treatment alternatives for localized prostate cancer are scarcely evaluated in Europe. Very limited available evidence supports the cost-effectiveness of radical prostatectomy over watchful waiting, brachytherapy over radical prostatectomy, and new treatment modalities over traditional procedures. Relevant disparities were detected among studies, mainly based on effectiveness. These apparently contradictory results may be reflecting the difficulty of interpreting small differences between treatments regarding QALY gains. Electronic supplementary material The online version of this article (doi:10.1186/s12913-016-1781-z) contains supplementary material, which is available to authorized users.


Background
Prostate cancer is the second most common cancer in men. An estimated 1.1 million men worldwide were diagnosed in 2012, with 345,000 cases in the European Union [1]. Estimates of public health expenditure on cancer indicate that prostate was the third contributor (6 % of the total), after colorectal and breast tumours [2]. Furthermore, United States (US) projections for the 2010-2020 period indicate a 27 % increase in cancer medical costs, where the largest is the continuing care phase of prostate cancer (42 %) [3].
The National Institute for Clinical Excellence (NICE) published a global systematic review of economic evaluations for localized prostate cancer treatments in 2003 [22], before the new surgical and radiotherapy modalities appeared. Since, only two other systematic reviews have been published on economic evaluations. One, focusing on radiotherapy [23], identified 14 studies. The other one, evaluating radical prostatectomy, did not identify any complete economic evaluation meeting inclusion criteria, but instead included 11 cost comparison studies [24]. To our knowledge, there is no global systematic review that takes into account the economic evaluations of all treatments published during the last 15 years, including those comparing different therapies, such as radical prostatectomy versus radiotherapy or active surveillance. As a consequence, the efficiency of existing treatment options for localized prostate cancer is still uncertain.
Most of the economic evaluations were conducted in the US [23][24][25][26], yet differences in health systems across countries limit their results' generalizability. Although there are also important differences within European countries, they share some major principles (such as a mainly publicly funded and almost universal coverage) far away from the insurance-based US health care system. Since economic evaluations are relevant to local context, our interest was centered in those performed in Europe. The aim of this study was to assess the efficiency of treatments in patients with localized prostate cancer, by synthesizing the available evidence from European economic evaluations through systematic review.

Methods
The protocol was registered in PROSPERO (http:// www.crd.york.ac.uk/Prospero) with number CRD42015 022063. We conducted systematic searches in MED-LINE, EMBASE and NHS EED (NHS Economic Evaluation Database, CRD York) databases with a specific strategy (see online Additional file 1) from January 1st 2000 to December 31st 2015.
We looked for economic evaluations (cost minimization, cost-effectiveness, cost-utility, and cost-benefit analyses) or cost comparison studies that assessed any modality of surgery or radiotherapy treatments, regardless of the comparator/s, for patients with localized prostate cancer (T1-T2). Articles were considered when referring to any European country, and published in any European language.
Studies were excluded if they only performed cost estimations without comparing treatments (such as cost studies, cost of illness studies, or budget impact analyses); they were not primary studies (reviews, editorials or commentaries); they assessed patients with advanced prostate cancer; or they evaluated diagnosis or screening procedures, but no treatments.
Two members of the study team (JJ and VB) independently reviewed articles found in the literature search by examining them in three consecutive phases: titles, abstracts, and full text. A third reviewer (MA) resolved discrepancies. A pilot test was performed to homogenize criteria among reviewers. Finally, the reference lists of the selected articles and those of previous systematic reviews were reviewed to identify other possible studies that could be included. Coding for inclusion and exclusion criteria were defined and recorded for each stage.
Assessment of studies' quality and data extraction was performed by the consensus of two reviewers (VB and MA). Drummond's Checklist was used for quality assessment [27]. Data was extracted using a standardized, prepiloted data collection form, including participant characteristics, interventions, comparator, economic perspective, and time horizon among others. The pre-defined primary outcome to be extracted was the incremental cost per Quality-Adjusted Life-Year (QALY) gained. Other Incremental Cost-Effectiveness Ratios (ICERs) and comparative costs per treatment were considered secondary outcomes. For illustrative purposes a figure has been designed to show all estimations of accumulated cost converted into euros (considering the current 2015 exchange rates), and plotted them through the time horizon for each intervention. Patient Intervention Comparator Outcome (PICO) strategy for this review is shown in the online Additional file 2.

Results
Literature flow in the systematic review Figure 1 shows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) diagram. Once 1,271 duplicates were excluded, 8,789 titles and 1,367 abstracts were reviewed, 165 articles were fully read, and finally only 13 eligible studies were included. Overall agreement and kappa coefficients (k) between reviewers were 79.7 % (k = 0.35), 92.8 % (k = 0.63), and 88.3 % (k = 0.53) in the title, abstract, and full text stages, respectively.
Most of the evaluations (nine out of 13) were conducted from a payer's perspective. Regarding the time horizon, lifetime (assuming an age limit of 100 years) was considered in five studies [22,28,32,36,37], one decade in three other studies [29,30,33], and shorter periods for the rest (from hospital stay to 5 years). Source of cost was medical records from study cohorts, such as the Scandinavian Prostatic Cancer Group Study Number 4 (SPCG-4) [40], or national database registers of activities such as the British National Health System (NHS) or, more rarely, only literature review (two studies) [36,37]. Similar sources were used for effects on health. Only in seven studies the threshold to consider an alternative as costeffective was clearly stated [28,29,32,33,36,37,41]. It ranged from €20,000 to €55,000 per QALY gained,    and four studies carried out sensitivity analysis around this threshold [22,28,32,33].

Main findings of economic evaluations identified in the systematic review
Estimated total direct cost for every treatment alternative was reported in all but two of the studies (see Table 2), which only showed incremental cost difference [29,37]. Eight studies could provide incremental cost per QALY gained [22,28,29,32,33,[35][36][37], and four studies other outcomes such as life year gained [28], 5year survival [31], successful treatment [35], and treatment side effects [39].
Of the interventions evaluated, three were found to be not only cost-effective but also dominant strategies (more effective and less costly): active surveillance over radical prostatectomy from a societal perspective in Germany [28], robotic-assisted over non-robotic surgical techniques [32], and IMRT over 3-Dimensional Conformal Radiation Therapy (3DCRT) when assuming a survival improvement of 6.6 years [36]. The following six interventions were found to be cost-effective: radical prostatectomy over watchful waiting in patients aged 70 or younger [29], robotic-assisted over non-robotic laparoscopic radical prostatectomy if more than 150 procedures performed per year [33], IMRT over 3DCRT when survival improvement is ≥3.8 years [36], and proton therapy [37], brachytherapy [22] and 3DCRT [22] over conventional radiotherapy. Conversely, the highest cost per QALY gained (least efficient options) were shown for radical prostatectomy versus watchful waiting in patients older than 75 [29], robotic-assisted versus non-robotic radical prostatectomy performing 50 procedures per year [33] (over £100,000), and for IMRT versus 3DCRT at equal doses and same survival to Prostate-Specific Antigen (PSA) progression [36] (over €100,000).
Estimations of accumulated direct costs in euros were plotted through the time horizon in Fig. 2 for each intervention. In total, the figure shows 38 estimates reported by 11 studies. The lowest costs (around €2,000) were obtained for expectant management (specifically, watchful waiting) at time horizons of 5 years and lifetime, as reported by Bauvin et al. [31] and Hummel et al. [22], respectively. The highest costs (around €24,000) were obtained for roboticassisted surgery during hospitalization [34] and for radical prostatectomy at 12 years [30].

Quality of the economic evaluations identified in the systematic review
The quality of the studies according to Drummond's 10item checklist is illustrated in Table 3. From the 11 economic evaluations, nine studies scored ≥8 points. The item that most frequently failed was about effectiveness, appraised uncertain or negative in six studies.

Discussion
Our systematic literature review identified 13 European studies, published 2000-2015, which conducted either economic evaluations or cost comparisons (11 and two, respectively) of any modality of surgical or radiotherapy treatments for localized prostate cancer patients. These studies varied widely in compared alternatives, costing methodologies, and time horizon. Estimations of incremental cost per QALY gained were provided by eight studies. Depending on the scenario and the comparator considered, three interventions were qualified as dominant (active surveillance [28], robotic-assisted surgery [32], and IMRT [36]), and six as cost-effective (radical prostatectomy [29], robotic-assisted surgery [33], IMRT [36], proton therapy [37], brachytherapy [22] and 3DCRT [22]).

Expectant management (active surveillance or watchful waiting) vs other treatments
Two cost-utility analyses comparing radical prostatectomy with expectant management show contradictory results: Koerber et al. [28] found that active surveillance was the dominant alternative (more QALYs at less cost), while Lyth et al. [29] showed that radical prostatectomy was more cost-effective than watchful waiting. However, the gain in QALYs in favor of active surveillance was extremely small (0.013) [28], and moderate-to-small in favor of radical prostatectomy (0.57-0.86) [29]. On the other hand, differences in the comparator used in both studies (active surveillance [28] and watchful waiting [29]) could also partly explain this disparity. No immediate treatment was performed in watchful waiting patients [29], while active surveillance involved [28] monitoring with PSA, digital rectal examination, and biopsy. Consistent with results reported by Lyth et al. [29], the cost-effectiveness study by Bauvin et al. [31] showed that radical prostatectomy is more effective than watchful waiting. Unfortunately, although the economic evaluation of Hummel et al. [22] also evaluated radical prostatectomy, they did not report its comparison with watchful waiting.

Robot-assisted laparoscopic prostatectomy (RALP) vs other surgical techniques
The previous systematic review of economic evaluations comparing robotic-assisted vs non-robotic laparoscopic surgery [24] proved to be insufficient for decision making, leading the authors to build a de novo economic evaluation [33], which has been now included in our review. Two of the three cost-utility studies that we identified consistently support the cost-effectiveness of robotic-assisted surgery [32,33]. Lord et al. [32] showed that robotic-assisted technique is the dominant alternative among surgery, while Table 3 Methodological quality assessment of economic evaluations using Drummond's 10-item checklist (Yes/no/can't tell) Koerber [28] Lyth [29] Bauvin [31] Hummel [22] Lord [32] Close [33] Hohwu [35] Hummel [36] Lundkvist [37] Becerra [38] Buron [39] 1. Was a well-defined question posed in answerable form? Number between square brackets corresponds to reference list position Close et al. [33] estimated a cost of £18,329 per QALY gained. Hohwu et al. [35] found no QALY gain for roboticassisted surgery, but the authors underlined the uncertainty of their QALY estimates due to a high degree of missing data. Again, disparity among these economic evaluations is mainly due to contradictory results on effectiveness, which were based on extremely small QALY gains for the roboticassisted technique: 0.007 reported by Lord et al. [32], and 0.08 by Close et al. [33] In fact, current guidelines of the European Association of Urology [5,6] consider all approaches (i.e., open, laparoscopic, and robotic) as acceptable for patients who are surgical candidates, because no single modality has shown a clear superiority in terms of functional or oncological results. On the other hand, it is important to highlight that the recommendation of the NICE Clinical Guideline [42] to provide robots in centers with an expected performance of at least 150 robotic-assisted operations per year, is only based on the economic evaluation published by Close et al. [33] It would be advisable to confirm this recommendation with future specific studies to help decision makers.

Conventional external radiotherapy vs new modalities
The systematic review of cost-effectiveness analysis by Amin et al. [23], comparing different radiation treatments, identified 14 studies (most from the United States, and only two from Europe [22,36]). Although evidence suggested that brachytherapy and IMRT were more costeffective than external beam radiotherapy, the authors highlighted the uncertainties and variation among studies [23]. We only identified three European economic evaluations comparing radiation therapies, each focusing on a different new modality (IMRT [36], proton therapy [37], and brachytherapy [22]). The three showed to be more cost-effective than conventional radiotherapy. However, each of these findings came from only one study, so further research is needed to confirm them. Once again, it is important to point out that the magnitude of the QALY gains is small for scenarios evaluating IMRT (0.01-0.613) [36] or proton therapy (0.297) [37], and moderate-tosmall in favor of brachytherapy (0.72) [22]. The European Association of Urology guidelines (5) recommend IMRT for definitive treatment with external radiotherapy, and brachytherapy for patients fulfilling specific criteria (low risk, prostate volume below 50 mL, no urinary obstruction, and no previous transurethral resection).

Prostatectomy vs radiation treatment
Of the three studies comparing prostatectomy with radiation treatment, only Hummel et al. [22] published a cost-utility analysis showing that brachytherapy was more cost-effective than surgery, with an incremental cost of €2,021-2,760 per QALY gained. Buron et al. [39] did not calculate ICERs but showed similar societal costs between radical prostatectomy and brachytherapy, though different treatment side effects: radical prostatectomy caused higher rates of urinary incontinence and erectile dysfunction, while brachytherapy presented irritative urinary and bowel symptoms more frequently. These results are consistent with the well-known side effect profiles of these treatments [8][9][10][11]. The costminimization published by Becerra et al. [38] assumed equal effectiveness in terms of survival, but did not take into account other relevant outcomes such as relapses and treatment side effects. Thus, evidence supporting the cost-effectiveness of brachytherapy over open radical prostatectomy originates from one single study [22] showing a small QALY gain (0. 35), and there are no economic evaluations comparing brachytherapy with robotic-assisted surgery.

Accumulated direct costs per treatment
As shown in Fig. 2, the cost-comparison study performed in Sweden reported the highest estimation of costs for radical prostatectomy and watchful waiting (€24,247 and €18,124) [30]; also, the cost-comparison study published by Barbaro et al. [34] showed an extreme perioperative cost in an Italian hospital for robotic surgery (€23,610). The high cost estimated in these two empirical cost-comparison studies [30,34] (based on the observation of health care activities in real cohorts) could indicate underestimation of real costs when they are based on models from theoretical cohorts. Furthermore, the surprisingly low accumulated costs estimated in most studies with theoretical cohorts and lifetime horizon [22,32,36], similar or even lower than those reported for studies with a shorter time horizon [31,33], also suggest an underestimation of real costs in these studies. Effectiveness is the most relevant component. However, the aforementioned disparities among studies in the identification of the most effective treatment may reflect the misinterpretation of such small QALY gains showed by the majority of them. For example, the gain of 0.013 QALYs [28] was much too small to consider active surveillance the dominant strategy over radical prostatectomy; or the gain of 0.007 QALYs [32] to consider robotic-assisted the dominant strategy over non-robotic techniques. Even the clinical relevance of the highest QALY gains identified in this review (0.57-0.86 for radical prostatectomy vs watchful waiting [29], and 0.72 for brachytherapy vs conventional radiotherapy [22]) may be questionable to be interpreted as relevant differences on effectiveness. Which is the reasonable cut-off for considering one intervention more effective than its alternative? Could gains lower than one QALY through 10 years or lifetime be considered clinically significant?

Cost and effectiveness components
Results from US economic evaluations [13,14] also showed no relevant differences in QALY gains for lifetime across treatments: ranging 0.5-1 or 0.7-0.8 for patients at low and intermediate risk, respectively, when comparing surgical and radiation therapies [13]; 0.9, 0.9, and 1.1 when comparing brachytherapy, IMRT and surgery with watchful waiting [14]. The clinical relevance of less than 1 year benefits between alternatives (in time horizons > 10 years of life) is questionable, and common sense prevents from interpreting them as differences in effectiveness.
An important issue related to the generalizability of study findings is the cost-effectiveness threshold, which represents society's willingness-to-pay for an additional unit of benefit [26]. Studies from UK showed a very consistent pattern regarding this threshold: they considered NICE's thresholds of £20,000-£30,000 per QALY gained [22,32,33,41]. Sweden studies showed a wider range for this threshold, from 200,000 SEK (€21,000) [29] to €55,000 per QALY gained [37]. The latter was very similar to the threshold applied in the German study (€50,000 per QALY gained) [28]. None of them was far from the US threshold's commonly accepted standard of $50,000 per QALY gained.

Limitations of the systematic review
There are several limitations that may affect our review findings. First, we cannot be sure that no relevant study is missing from this systematic review. However, in order to find as many relevant studies as possible, we have performed the search in PubMed and EMBASE, the most comprehensive databases in health sciences, as recommended [43], as well as in a specific database for economic evaluations. In addition, we designed a very sensitive search strategy (yielding the 8,789 titles revised) and we performed an additional manual reference search. Second, no quantitative synthesis of the results by meta-analysis was planned due to the well-known high heterogeneity among health economic evaluations. Furthermore, considering the scarce number of studies comparing the same interventions, obtaining a pooled estimator would make no sense. Third, internal validity of the synthesis provided by a systematic review depends on the quality of primary studies. In our systematic review, quality could be considered good except for effectiveness, which failed in almost half of the studies. It is necessary to take into account that recruitment for randomized trials presented considerable difficulties in these patients [44,45], and the only available trial, the SPCG-4 [40]-which was used in several of these economic evaluations, was conducted at the beginning of PSA era. Fourth, studies with a cost-comparison design were included despite not being economic evaluations. However, the information they provided clearly contributed to the amount and robustness of evidence on costs.
Finally, Fig. 2 shows reported direct healthcare costs without transforming them into a single year to avoid manipulation. We only converted currency into euros, using 2015 exchange rates, to facilitate comparisons.

Conclusions
To our knowledge, this is the first systematic literature review of the European economic evaluations of all main primary treatments for localized prostate cancer published during the last 15 years. The 13 studies identified (five comparing interventions with expectant management, four contrasting robotic with non-robotic assisted surgery, three assessing new modalities of radiotherapy, and three comparing radical prostatectomy with brachytherapy) showed that currently relevant treatment alternatives for localized prostate cancer are scarcely assessed in economic evaluations in the European countries. Furthermore, differences between cost-comparison and cost-effectiveness studies suggest underestimation of costs in studies based on models from theoretical cohorts.
In conclusion, very limited evidence supports the costeffectiveness of radical prostatectomy versus watchful waiting, and that of brachytherapy versus radical prostatectomy. Regarding the evaluation of new treatment modalities, also limited evidence supports the cost-effectiveness of roboticassisted laparoscopic radical prostatectomy versus nonrobotic procedures, and that of brachytherapy, IMRT and proton therapy versus traditional external radiotherapy. Relevant disparities were detected among studies, mainly based on effectiveness. These apparently contradictory results may be reflecting the difficulty of interpreting small differences between treatments regarding QALY gains. Moreover, despite an acceptable methodological quality in most aspects of the studies included, the effectiveness