Ablative therapy for people with localised prostate cancer: a systematic review and economic evaluation

Importance of prostate cancer as a health problem

Prostate cancer is the commonest cancer diagnosed in people in the UK and is the second commonest cause of cancer deaths. 1 In 2011, 41,736 people in the UK were diagnosed with prostate cancer (Figure 1). 2 It accounts for approximately 7% of cancer-related deaths in people in the UK, with an age-standardised mortality rate of 35 in 100,000, amounting to 10,837 people in 2012. 2 In 2006, the 10-year prevalence in the UK was estimated to be 181,463. 3 The disease also incurs significant economic costs to health-care providers. In 1997 the annual cost to the NHS was estimated at £55M. 4 An economic analysis published in 2012 reported that the total cost of prostate cancer in the UK in 2009 – encompassing treatment costs for surgery, radiotherapy, hospital and community care, premature deaths, time off work and unpaid care given to patients by family and friends – was estimated at around £800M per annum. 5 This ties in with a recent estimate from the USA that a diagnosis of prostate cancer incurs an average health-care and personal cost of US$20,000 (£13,000) over the individual’s remaining lifetime. 6,7

FIGURE 1.

Age-specific incidence rates for prostate cancer in the UK, 2009–11. 3

Since the advent in the mid-1980s of testing for prostate-specific antigen (PSA), an organ-specific serum marker of prostate cancer, there has been a substantial increase in the number of people diagnosed with prostate cancer. 8 The largest rise in incidence is among relatively younger people as a consequence of case-finding and screening for asymptomatic disease using serum PSA and multiple transrectal ultrasound (TRUS)-guided needle biopsies of the prostate. 9,10 By the same token, the use of PSA testing has resulted in a gradual stage migration towards the earlier stages of the disease in terms of diagnosis. Indeed, presently the majority of people (i.e. up to 80%) with prostate cancer are diagnosed in the localised stages of the disease,11,12 and a large proportion of them often have favourable pathological characteristics. 13,14

Standard curative treatment options for localised prostate cancer

The decision to treat and the choice of treatment are influenced heavily by four major factors:

1. the patient’s life expectancy, as determined by his chronological age, comorbidities and fitness in terms of activities of daily living (called performance status)

2. tumour characteristics, determined by the PSA level at diagnosis; the aggressiveness of the tumour, as determined by histological examination using a microscope [or tumour grade, categorised by Gleason sum score (2–10)]; other histological parameters, including volume of the cancer and length of involvement of the biopsy strands of tissue with cancer; and the stage (or extent of spread) of the disease on clinical examination and imaging, all of which can be used for risk stratification to predict behaviour or outcomes using nomograms15,16

3. clinician or patient preference linked to risk of adverse effects

4. availability of resources underpinning each treatment option.

A standard clinical treatment care pathway for people with localised prostate cancer is given in MacLennan and colleagues,17 and this is further described in Chapter 2. As the majority of people present with asymptomatic, early and localised disease, most of them can be cured by way of radical (or curative) treatment options, which include either RP or radical EBRT. However, it is also clear that prostate cancer has a wide spectrum in terms of the risk and time course of disease progression,18 and in spite of the use of nomograms, the disease course for some patients can be unpredictable.

Evolution of ablative therapies for localised prostate cancer

Ablative therapies refers to a group of interventions which aim for either total, subtotal or focal ablation (or destruction) of the prostate gland in order to treat localised prostate cancer. These therapies have some common characteristics, including (i) a minimally invasive nature, that is they are performed percutaneously through the perineum, transurethrally or transrectally; (ii) being technically simple and easy to master; (iii) allowing repeat treatments; and (iv) allowing salvage radical treatment for treatment failure (i.e. failure to eradicate disease) or disease recurrence following initial cure. These therapies destroy the cancer in the prostate gland in a minimally invasive manner using a range of energy sources, while simultaneously minimising damage to adjacent structures such as the urinary sphincter, urethra, bladder, rectum and nerves for erectile function, hence potentially reducing the risk of adverse effects.

The technology was first described in the mid-1990s, with cryotherapy, high-intensity focused ultrasound (HIFU) and interstitial brachytherapy being used to treat localised and locally advanced prostate cancer in a non-focused manner, whereby the entire prostate gland was subjected to treatment. Over the past two decades, advances in imaging by ultrasound (US) or MRI, together with template biopsy protocols and improvement to the treatment technologies, have all driven the possibility that these therapies can be used in a more precise way, whereby the part of the prostate harbouring the most aggressive cancer can be preferentially targeted for destruction. 43,44 This is achieved in several ways, including lesion-targeted therapy, hemi-ablative therapy (where one half of the gland is targeted) and subtotal ablative therapy. This approach enables preservation of areas of the gland without cancer, together with surrounding structures such as the nerves and blood vessels for erectile function, and the urinary sphincter muscle, bowel and bladder, hence potentially reducing the risk of adverse effects. The approach is also potentially applicable to the common finding of multifocal disease, where the dominant foci with less favourable pathological characteristics are treated, while other, smaller, low-risk foci are left and AS continued. 44 Although primarily undertaken using general anaesthetic, ablative therapies may also be performed under local anaesthetic or sedation in the outpatient setting. Other advantages include the ability to repeat the ablative procedure if required, and if the procedure fails to achieve cancer control, then salvage treatment by way of surgery or radiotherapy can be undertaken.

In addition, ablative therapies have also been used in treating people with local relapse after radical EBRT. Although radical EBRT is considered a curative treatment option for localised prostate cancer, a relatively high proportion of people, estimated at approximately 30%,45 will develop recurrent disease signalled by a rising PSA and a positive rebiopsy. This recurrence rate is, to some extent, inflated by the higher proportion of people being treated for more advanced disease compared with RP. If left untreated, at least 75% of these people will develop localised prostate cancer recurrence within 5 years46 and hence require further treatment, although the timing of second-line treatment remains controversial. Subsequent treatment options include palliative hormonal therapy and potentially curative salvage procedures. The currently recommended option, salvage prostatectomy, carries a high risk of morbidity including urinary incontinence and rectal injury.

The ablative technologies considered in this review are (1) brachytherapy; (2) cryotherapy; (3) HIFU; (4) vascular-targeted photodynamic therapy (PDT); (5) transperineal radiofrequency interstitial tumour ablation (RITA) therapy; and (6) laser ablation therapy (encompassing procedures such as photothermal therapy, laser interstitial tumour therapy and laser photocoagulation).

Technical description of the interventions

Projected rise in the number of people in the UK requiring treatment for localised prostate cancer

At present in the UK, localised prostate cancer is detected by case finding among asymptomatic people who request a PSA test and during the assessment of people complaining of unrelated urinary symptoms. In 2010, almost 41,000 people were diagnosed with prostate cancer in the UK,1 with 18,408 (45%) aged younger than 70 years. 3 The majority of these people will have localised-stage disease, and are hence suitable for curative treatment. 1 Previous annual estimates of treatment suggest that over a 12-month period, 3922 people underwent RP,64 while approximately 4000 underwent EBRT and 1455 underwent brachytherapy. 64 The corresponding figure for AS was approximately 800. 65 There is evidence from the USA to suggest that the use of RP as a primary treatment option for localised prostate cancer is increasing. 34 The results from a PSA screening trial, the European Randomised Study of Screening for Prostate Cancer (ERSPC), showed a doubling of cancer detection rate among people in the target age group (55–69 years) accompanied by a similar increase in the number of people going on to have potentially curative treatment. 66 Overall, 3% of people screened and 1% of controls underwent RP during the 9 years of follow-up. Findings from the US-based Prostate, Lung, Colon and Ovarian Cancer Screening Study (PLCO) were similar, with 3% of people screened and 2% of controls having RP during the 10-year study duration. 67 Translating these figures to the UK 2011 population of 5.05 million people aged 55–69 years, the annual number of RPs would rise to approximately 7000 with increased case finding and to 11,000 if a screening programme was instituted.

Recent evidence from the HTA-funded UK trial of treatment for localised prostate cancer, Prostate Testing for Cancer and Treatment (ProtecT), suggests that the incidence of disease in younger people aged under 55 years is also significant, further increasing the population potentially requiring consideration towards treatment. 68 Evidence from the USA suggests that increasing incidence of low-risk cancer is accompanied by increased use of AS and newer ablative therapies such as cryotherapy, with AS being selected by 10.2% and newer ablative therapies by 4.4% of affected people between 2004 and 2006. 34 In NHS hospitals in England, the numbers of people with prostate cancer treated with newer ablative options remains small, with 66 recorded as undergoing cryotherapy and 168 HIFU,64 although discussion with relevant clinicians suggests that the numbers are increasing.

Introduction

The population of patients considered for this review are people with localised prostate cancer who are considered suitable for active treatment or AS and are managed within the UK NHS. The patient characteristics that define this population include age and comorbidity that collectively determine an estimated life expectancy of at least 10 years.

Disease factors provide the estimated risk of developing recurrent disease, either from distant metastases not identified at pre-operative assessment, or because of failure to completely remove localised disease. The approximate magnitude of this risk for an individual diagnosed with prostate cancer can be calculated using a nomogram. The most commonly used version is hosted by the Memorial Sloan Kettering (MSK) Cancer Institute in web-based form. 69 These models use the pre-operative disease variables of age, PSA, clinical tumour stage, Gleason grade and number of needle biopsy cores positive for cancer.

The described care pathway was constructed using available evidence, previous care pathways (Figure 2) developed by the Aberdeen Academic Urology group in conjunction with a national and international panel of experts, and consensus-building through several meetings of the expert panel convened for this review. Although it is primarily constructed as the basis of the modelling of cost-effectiveness reported in Chapters 9 and 10, the pathway is consistent with previously published clinical pathways of care. 9,70–73 The complete care pathway developed for the review is shown in Chapter 9 (see Figure 16, with Figures 17–20 illustrating how the care pathway varies for alternative interventions under investigation).

FIGURE 2.

Localised prostate cancer care pathway. 17

Pretreatment level of prostate-specific antigen

The pretreatment PSA level is an independent statistically significant predictor of future recurrence, but on its own is limited in reliability and predictive value. For prognostic purposes the value is defined in risk groupings corresponding to low (< 10 ng/ml), intermediate (10–20 ng/ml) and high (> 20 ng/ml) risk of disease progression following radical treatment. 36

Staging of prostate cancer

The stage of an individual’s cancer is categorised according to the Union for International Cancer Control (UICC) 2009 tumour node metastasis (TNM) classification. 74 Pre-operatively, this is determined by clinical assessment using DRE and imaging and is given the prefix ‘c’. Following removal and pathological examination of the prostate and, in some cases, adjacent lymph nodes, the staging is adjusted accordingly and given the prefix ‘p’.

Gleason grading

The qualitative low-magnification microscopic histological description of prostate cancer first suggested by Gleason in 196675 remains an essential aspect of prognostic categorisation, although there have been substantial modifications over the years. 76 Standard practice consists of identifying the first and second most prevalent patterns within a set of biopsy cores which give the primary and secondary Gleason grades (each rated 1–5). These are then added together to give the overall Gleason sum score (2–10). Recent consensus tends to limit the use of grades 1 and 2 and therefore scores generally range between 6 and 10. 77 Higher individual grade and total score indicate more aggressive disease, with primary grade being more predictive. An individual whose tumour is categorised as Gleason score 4 + 3 = 7 will therefore tend to have a worse prognosis than if the Gleason score was 3 + 4 = 7, for example. 78

Cancer volume

There is some evidence that cancer volume is also an independent prognostic factor for progression of the cancer following initial management. For this reason, pathologists examining biopsy cores will estimate cancer volume by stating the number of cores that contain cancer and estimating the proportion of each core that is affected by the cancer. 79

Summary

Pretreatment information including age, serum PSA, tumour stage, Gleason sum score and tumour volume predicts the risk of disease recurrence. It is therefore important that studies comparing treatments, such as this current assessment, include an evaluation of whether or not the patient groups undergoing each procedure are balanced for these variables. For the purposes of the current assessment, people with localised prostate cancer will be stratified into three groups according to D’Amico risk of recurrence following curative treatment36 (Table 1): low, intermediate and high risk. The system utilises pretreatment variables of serum PSA level, Gleason sum score and T stage of the TNM staging system.

Introduction

This study includes a cost-effectiveness analysis of ablative therapies compared with other standard interventions (see Chapters 9 and 10 for more details). For the economic modelling, it is assumed that the procedures being considered will be carried out in hospitals that have the necessary resources in terms of staff, facilities and NHS cancer plan approval to carry out the various interventions on a routine basis. For surgical procedures (i.e. RP, HIFU or cryotherapy), this will comprise operating theatre and recovery facilities, including critical care and standard urology wards; the required clinical and technical expertise, including surgeons, anaesthetists, theatre nursing team, pathologists and technicians; and continued care, including outpatient review, repeat imaging and facilities for further treatment for adverse events or cancer progression. For EBRT, the resource estimates were modelled after IMRT, because in most cancer units around the UK, IMRT has superseded 3D-CRT as the standard for EBRT. The resource estimation includes costs associated with radiotherapy planning visits, treatment sessions, staff time, consumables, etc. For brachytherapy, the resource estimates were modelled after low-dose brachytherapy (i.e. involving permanent seed implantation), and resource estimation includes costs for seed implantation under general anaesthetic, incorporating costs for a radiologist, urologist, oncologist, anaesthetist, theatre staff, consumables, etc.

A detailed description of the various interventions are provided in Chapter 1, and a more detailed description of the treatment care pathways for each intervention in terms of resource use is provided in Chapter 9.

Learning curve of procedures

For safe conduct of all interventions, it is essential that all members of staff involved in delivering the therapy have had specific training and are competent to undertake the procedure. In the UK, all of the surgical procedures (including RP, cryotherapy and HIFU) are normally performed by consultants who have received specific training. As such, for the economic modelling, it is assumed that such procedures are undertaken by a trained consultant. Non-surgical procedures such as EBRT and brachytherapy are less susceptible to learning curve effects of individuals. The review assumes that such procedures are undertaken by experienced teams led by a consultant oncologist.

Hospital stay

For surgical procedures, people are generally admitted to hospital either on the day of surgery or the evening before. For RP, a rectal enema is administered to clear the lower bowel. Immediately prior to the procedure, prophylactic antibiotics are given according to local policy and venous thrombosis/embolism prophylaxis is commenced as required. After surgery, the patient is routinely nursed on a standard ward although specific comorbidities or intraoperative complications may require a period in a critical care bed. For RP performed laparoscopically, people are typically discharged home after 3 days with an indwelling catheter, although this may be variable (e.g. hospitalisation time can be reduced by managed care programmes). They then return to the ward as a day patient after a further 7–14 days, according to local protocol, for urinary catheter removal and voiding check. For cryotherapy, people stay up to 2 nights in hospital after their procedure, whereas for HIFU they stay for 0–1 night. In both instances, they return to the ward after a further 7–14 days as a day patient for urinary catheter removal and voiding check.

Perioperative complications

Although people undergoing surgery for localised prostate cancer (including RP, cryotherapy and HIFU) generally do not have concurrent comorbidity that is a persistent threat to their health, a proportion will be expected to suffer adverse events associated with surgery, and anaesthetic-related problems such as cardiac ischaemia and pulmonary embolism. In addition, specific complications include urinary and blood stream infection, inadvertent injury to adjacent organs (e.g. rectal injury), excessive blood loss requiring transfusion, prolonged urinary or lymphatic leakage from abdominal drains, development of urethral stricture or fistula, etc. The adverse effect of these complications in terms of their severity and requirement for additional interventions and hospital stay can be summarised according to the Clavien system (Table 2). 80,81

For RP, an additional specific short-term complication is narrowing (bladder neck stenosis) of the sutured join between the top of the urethra and bladder outlet (vesicourethral anastomosis). This will become noticeable after removal of the draining catheter and will result in voiding problems reported by the patient at the 6-week outpatient review. It is treated with endoscopic incision of the narrowed area, which requires an additional short hospital stay and a 7-day period of catheterisation. For most people the problem is cured by a single incision, although for some this may need to be repeated once or twice. 82

For non-surgical interventions (e.g. EBRT and brachytherapy), the management of adverse events depends on the severity, graded according to common acute and late toxicity grading systems [e.g. Radiation Therapy Oncology Group (RTOG) Common Toxicity Criteria]. 83 In this assessment, for the estimation of resource use, expected duration of hospital stay based on the severity of adverse events was graded based on clinical judgement from members of the study team.

Histopathological examination of prostate biopsies and radical prostatectomy specimen

For RP, careful and thorough microscopic examination of the removed prostate by an experienced pathologist is required to determine the true extent of the disease, to decide whether or not the surgery may have been unable to remove all the contained cancer (positive margin) and whether or not the cancer had spread outside the prostate (extracapsular extension), and, if lymphadenectomy has been performed, to detect the presence of lymph node metastatic disease. In addition, a more comprehensive description of the distribution of Gleason patterns within the cancer is possible. This examination will recategorise the disease according to stage [pathological tumour (pT) and pathological node (pN)] and postoperative Gleason score, which will allow more accurate estimation of prognosis according to available post-RP prognostic nomogram69 and inform whether or not early additional (adjuvant) treatment should be advised. The crucial nature of this examination has led to consensus meetings of expert pathologists who have set out a specified protocol of specimen collection, processing, examination and analysis. 77,84

For interventions whereby repeat prostate biopsies are necessary as part of the follow-up protocol (e.g. AS, cryotherapy and HIFU), or triggered by a suspicion of biochemical recurrence, the economic model assumes that the biopsies are performed using the TRUS approach, and where appropriate, this may be augmented by MRI-directed or guided strategies. The biopsy specimens are reviewed and reported by an experienced pathologist within a urological cancer multidisciplinary team setting.

Follow-up schedule

People who have undergone RP are generally seen by the operating team as outpatients 6 weeks after their surgery, then 3-monthly for the first year and 6-monthly for the next 4 years. At each follow-up, serum PSA is checked for tumour recurrence and a qualitative assessment made for continence and desired sexual function. If further assessment or treatment is required for any of these aspects, then the pathway of care will be changed accordingly.

For EBRT and brachytherapy, patients were assumed to have follow-up as part of post-treatment surveillance for up to 5 years, assuming that there were no changes in the patient’s condition, nor any evidence of biochemical recurrence such that they had to leave the surveillance state. In the first year of surveillance, patients would attend four nurse-led urology outpatient appointments, with PSA tests conducted in each of these. For the second year through to the fifth it was assumed that patients would attend two nurse-led urology outpatient appointments with PSA testing at each. After the first 5 years, it was assumed that patients would receive an annual PSA test conducted by a practice nurse in a primary care setting. Patients would also have an annual DRE each year for the first 5 years.

For cryotherapy and HIFU, similar assumptions were made. Where repeat prostate biopsies were mandatory within the first year of treatment, this was regarded as a part of the ‘package of treatment’ rather than as a part of follow-up. The economic model also made allowance for imaging of the prostate using multiparametric MRI during the follow-up period.

For AS, the following assumptions were made based on a standard protocol. In the first year of follow-up, patients would attend four nurse-led urology outpatient appointments with PSA tests conducted at each appointment. At 12 months, a multidisciplinary team cancer meeting would take place to review each patient. Patients in year 2 would attend two nurse-led urology outpatient appointments, again with PSA tests performed at each appointment. In addition, patients would undergo a standard 12-core TRUS-guided biopsy. Year 4 of AS was assumed to be identical to this, and years 3 and 5 were assumed to be the same with the exception of the TRUS-guided biopsy. Patients would also have an annual DRE in years 1–5. After the first 5 years, it was assumed that patients would receive an annual PSA test conducted by a practice nurse in a general practice setting.

Detection of persistent or recurrent disease

The risk of disease recurrence is higher if one or more of the disease factors, including pre-operative PSA > 20 ng/ml, Gleason score of > 7, extracapsular disease (T3/T4), positive margin or positive lymph nodes (N stages N1/N2 of the TNM staging system), are present as determined by lymphadenectomy. If the likelihood of disease persistence or recurrence is deemed to be very high, then immediate adjunctive treatment may be offered. For the majority of people, PSA surveillance is started according to the above schedule. There are multiple definitions of the threshold of PSA rise that signifies biochemical recurrence between interventions, and within an intervention. For RP, because the prostate gland and prostate cancer (which are the only sources of PSA in the blood) have been removed, if cure has been achieved, the expectation is a complete absence of serum PSA 3 weeks after treatment. However, laboratories have different sensitivity and specificity thresholds. The commonest baseline is 0.2 ng/ml. As such, for a definition of cure, the patient should have reached a nadir (i.e. lowest PSA reading) which is below 0.2 ng/ml after 3 weeks following treatment. The most common definition for biochemical recurrence is two successive serum PSA readings > 0.2 ng/ml. 85

For EBRT and brachytherapy, several definitions are in existence, the commonest of which is the Phoenix definition. 86 This defines recurrence as ‘a rise by 2 ng/ml or more above the nadir PSA’.

For cryotherapy and HIFU, there is no consensus regarding what should constitute biochemical recurrence. Although the Phoenix criterion is often reported, it has not been validated for either intervention.

For AS, because the cancer remains untreated but merely monitored, definitions for biochemical recurrence do not apply. The main immediate cancer-related outcome of relevance for AS is disease progression or upgrading of cancer grade (often collectively termed ‘reclassification of disease’). However, there is controversy regarding what constitutes progression or reclassification, with AS protocols adopting different definitions. 87

For all interventions except AS, the occurrence of biochemical recurrence does not automatically trigger salvage treatment; in some instances, the patient may continue to be monitored until a point where salvage treatment is deemed necessary. However, most patients will undergo salvage treatment once biochemical recurrence occurs. The decision whether to institute immediate salvage treatment or further monitoring will be informed by tests such as MRI and/or a radionuclide bone scan designed to demonstrate the site of recurrence as being in the prostatic bed (i.e. localised recurrence), or as lymph node or bone metastases (i.e. systemic recurrence).

Salvage treatment

Following localised recurrence, the salvage treatment options are salvage RP, salvage EBRT, savage brachytherapy and salvage ablative therapy (HIFU or cryotherapy) (see Figure 16 in Chapter 9). The assumption for the model is that salvage treatment should differ from the primary treatment (i.e. patients who have had primary RP would be ineligible for salvage RP). With the exception of salvage RP, the model allows for the addition of androgen deprivation therapy for a duration of up to 2 years following salvage treatment.

For people with likely systemic recurrence, long-term androgen deprivation therapy (medical castration), most commonly achieved with a luteinising hormone-releasing hormone (LHRH) agonist, is recommended. This consists of 3-monthly injections of a depot preparation of the chosen drug. For people whose disease progresses despite local and systemic adjuvant treatment, palliative symptom control will be instituted.

Urinary incontinence

Urinary incontinence is one of the most important long-term adverse effects of treatment for localised prostate cancer. Recovery of continence following some interventions, such as RP, can take up to 12 months, although most people will regain continence by 6 months. Therefore, people suffering urinary incontinence will be advised to use containment devices such as absorbent pads or penile sheath drainage for the initial 12 months. For the majority of interventions, the expectation is for urinary incontinence to improve within the first year, beyond which further improvement is unlikely. As such, if bothersome leakage persists beyond this time, then the main treatment options will be surgical implantation of an artificial urinary sphincter (AUS) or continued use of containment devices.

Erectile dysfunction

Of people who were sexually active prior to treatment, a large proportion will experience worsening of their sexual function, and in particular difficulty initiating and sustaining penile erection sufficient for intercourse. This is particularly dependent on preservation of one or both neurovascular bundles during treatment. For these people, treatment options will include drug treatment taken as required, vacuum constriction device or penile implant surgery. Most people will first trial the oral phosphodiesterase inhibitors sildenafil, tadalafil (Cialis^®, Lilly) or vardenafil (Levitra^®, Bayer) which, under NHS prescribing rules, are limited to one tablet weekly. The next option will be alprostadil given as an intraurethral pellet or an intracavernosal injection with NHS supply, again limited to once-weekly doses. For people who achieve satisfactory restoration of sexual activity with these drugs, it is assumed that their use will continue long term. If drug treatments are unsuccessful, people may trial a vacuum constriction device, or consider surgical implantation of a penile prosthesis. The proportion of people pursuing the last two options is small, as most will accept their loss of sexual function in the longer term.

Identification of other relevant information, including unpublished data

The World Health Organization (WHO) International Clinical Trials Registry, EU Clinical Trials Register, Current Controlled Trials, ClinicalTrials.gov and National Institute for Health Research (NIHR) Portfolio were searched for ongoing studies. Websites of manufacturers, professional organisations, HTA organisations and regulatory bodies were also checked for additional reports (see Appendix 1, Websites consulted).

Types of study

For all three reviews, we considered evidence from RCTs and non-randomised comparative studies (NRCSs) (if no RCT evidence was identified), and from single-arm cohort studies (case series) (greater than 10 participants) for the ablative procedures only. Had comparative studies of the ablative procedures been identified, consideration would have been given to removing single-arm cohort studies from the reviews.

Studies comparing only multiple treatments of the same non-ablative therapy within the same comparative study (e.g. comparing different dosages of radiotherapy, or open vs. laparoscopic prostatectomy) were excluded. Conference abstracts were excluded, as were non-English-language reports with the exception of RCTs incorporating an ablative procedure comparison, for which no language restriction was applied.

Types of participants

The types of participant considered were people with localised prostate cancer, defined as cancer confined to the prostate gland. Eligible patients had clinical stage T1 or T2 disease at presentation (not pathological staging).

We planned to stratify people into localised low/intermediate risk and localised high risk of progression, based on the criteria shown in Table 3 (adapted from D’Amico risk stratification). 69

The criteria for assessing the patient’s risk of recurrent disease were the same for primary or salvage treatment. For studies with patients of mixed clinical stages (i.e. T1 to T4), studies were included if greater than 80% of the patients were stage T1 or T2. Additionally, for the salvage therapy review the patients must have received EBRT prior to salvage therapy being considered. Studies of people with locally advanced prostate cancer (considered as stage T3/T4) were excluded.

Although the systematic reviews of primary treatment of localised low-/intermediate-/high-risk prostate cancer and salvage therapy relate to subsets of T1 and T2 disease, we included any studies that reported comparative data on T1 and/or T2 disease. This reflects the observation during scoping (and our experience of conducting such reviews in prostate cancer) that many studies do not report outcomes by the substages of T1 or T2 disease. Given the difficulty in attributing studies to subsets of T1 and T2 disease, it was not possible to undertake analyses of subsets on risk.

For the primary review, studies were included if patients were fit for surgery. Where studies enrolled patients for both primary and salvage procedures and reported combined results, the study was eligible if 5% or less of the study population were salvage patients.

For the salvage review, studies were included if at least 80% of all salvage patients had received prior treatment with EBRT.

Types of interventions and comparators

For the primary therapy systematic review on low-/intermediate-risk localised prostate cancer, the ablative therapies considered were cryotherapy, HIFU, PDT, RITA, laser ablation and brachytherapy. The comparators were AS, RP and EBRT.

For the primary therapy systematic review on high-risk localised prostate cancer, the ablative therapies considered were cryotherapy, HIFU, PDT, RITA, laser ablation and brachytherapy. The comparators were RP and EBRT.

For the salvage therapy systematic review, the ablative therapies considered were cryotherapy and HIFU. The comparator was RP.

Types of outcome measures

In addition to contacting content experts to identify outcomes of importance, we also elicited the views of a group of people living with prostate cancer. The group consisted of seven male participants who had undergone ablative therapy (HIFU), robotic, laparoscopic and open RP, and radiotherapy. The participants were invited to join a group discussion and express their own opinions on the choice of relevant outcomes following treatment for localised prostate cancer. They were recruited through a local Urological Cancer Charity (UCAN) and were not aware of the views of the content experts.

On the whole, the participants were in agreement with the content experts as to the key outcomes of importance. For example, clear primary importance was placed on survival and recurrence (cancer-specific outcomes). Several people commented that other outcomes were irrelevant in the event of death. Survival was deemed the most important outcome, but some noted that, in the context of localised disease, they assume that they will survive the cancer and so other outcomes then take on more importance. The interaction between survival, recurrence, progression and treatment success was also considered important in treatment decision-making.

Other outcomes that were highlighted as being meaningful to all of the participants were urinary incontinence and erectile dysfunction, followed by quality of life. Outcomes that were mentioned by some of the participants were catheterisation, urethral stricture, Peyronie’s disease, length of hospital stay, faecal incontinence, rectal itching and bleeding, emptying the bladder when ejaculating, having to travel for treatment, getting ‘back to normal’ and recovery times. Financial cost to the NHS was not deemed to be of high importance.

The outcomes considered in this assessment were categorised as follows:

• cancer related

  • biochemical (PSA) recurrence (primary cancer-related outcome)43

  • disease-free survival, defined as the absence of clinically detectable disease in a surviving patient

  • overall survival

  • further prostate cancer treatment

• adverse effects: functional outcomes

  • sexual (penile erection) function, defined by validated score [such as the International Index of Erectile Function-5 (IIEF-5)], or as defined by the triallists

  • urinary continence, defined, for example, as ≤ 1 thin pad per day and/or by validated symptom score [such as the International Consultation on Incontinence Modular Questionnaire – Urinary Incontinence (ICIQ-UI)], or as defined by the triallists

• quality of life

  • generic and disease-specific quality of life [validated quality of life score such as the Short Form questionnaire-36 items (SF-36)]88

• procedural

  • length of hospital stay (if applicable)

  • abandonment of the procedure

• adverse events: procedural complications and early death

  • including, but not restricted to, urethral sloughing, rectourethral fistula formation, urethral stricture formation, acute urinary retention, dysuria, pelvic pain, rectal injury, perioperative death, and periprocedural death and Clavien score (if applicable).

Exclusion criteria

The following types of report were excluded:

• reports focusing on people with metastatic disease

• non-English-language reports of non-randomised studies

• conference abstracts

• reports of retrospective studies of AS.

Risk of bias

Experience has demonstrated that multiple quality assessment tools are required for systematic reviews where multiple study designs are considered. One reviewer assessed the quality of included studies using one of three prespecified checklists, depending on study design. The standard Cochrane risk-of-bias tool89 was used to assess the risk of bias in randomised trials, and the risk-of-bias tool recommended by the Cochrane Non-Randomised Studies Methods Group was used for NRCSs. 89 For NRCSs, the main confounders were identified a priori by the expert panel (by outcome). A study was considered to be at high risk of bias if any of the confounders were imbalanced (e.g. age or D’Amico risk). We developed a case series tool for assessing risk of bias through our partnership in the Review Body for Interventional Procedures for NICE. 90–93 The case series tool rates bias and generalisability, sample definition and selection, description of the intervention, outcome assessment, adequacy of follow-up and performance of the analysis. Discrepancies were resolved by discussion or referred to a third party. Copies of the risk-of-bias tools are given in Appendices 3–5.

Pre-planned subgroup analyses

After discussion at the first expert advisory group meeting for this assessment, three subgroup analyses were identified to be undertaken. The subgroups were:

• low risk of bias studies only

• focal ablative therapy versus EBRT or RP

• low-risk disease treated with ablative therapy versus AS.

Number of studies identified

Title and abstract searches identified 7134 potentially relevant citations, from which 548 reports were retrieved for full-text screening. Of these, 121 were included and 427 were excluded, with reasons given in Figure 3. Of the 121 included reports, 113 reports (88 studies) were eligible for inclusion in the review of patients undergoing primary treatment for localised prostate cancer,36,49,52,98–207 and eight additional reports (nine studies) were included in the review of patients undergoing salvage treatment for recurrence of local prostate cancer following EBRT failure. 208–215 In one report,120 two studies were reported, each eligible for primary and salvage reviews; another121 reported data separately for a subset of participants who were randomised and also for all randomised and non-randomised participants. These reports were treated as contributing two studies each to this review. Uchida and colleagues195 was found to be related to five other reports;183,190,192–194 Ferrer and colleagues130 was related to two reports;137,167 Shah and colleagues182 was related to Mohammed and colleagues164 and Vicini and colleagues;201 Bul and colleagues111 was related to van den Bergh and colleagues;197 Klotz 2010146 was related to three reports;147,148,157 Selvadurai and colleagues181 was related to van As and colleagues;196 Caso and colleagues114 was related to a secondary report with the same lead author115 and to Polascik and colleagues;175 Truesdale and colleagues188 was related to Lambert and colleagues;152 Donnelly and colleagues125 was related to Robinson and colleagues;179 another study by Donnelly and colleagues124 was related to three reports;177,178,180 and Paulson and colleagues168 also published their report in German. 169 Appendix 6 details the references of the included reports and shows the linked reports, and Appendix 7 details the excluded reports.

FIGURE 3.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram of potentially relevant reports of identified studies and the numbers subsequently included and excluded from the clinical effectiveness review. a, one included in both reviews.

Number and types of studies included

Of the 88 studies, four RCTs were included in the primary review, one each comparing cryotherapy with EBRT125 and EBRT with RP,168 and two comparing brachytherapy with RP. 49,121 Forty NRCSs,36,100,101,103,105,108–110,113,117,119,121,123,126,128,130,131,135,136,144,145,149,151,153,156,160,163,170–172,176,182,184,186,189,198,203,205–207 including 25 prospective studies,100,101,103,108–110,113,117,121,123,128,130,144,145,149,153,156,160,163,172,176,182,184,186,198 were included in the primary review. The method of data collection could not be determined for the study by Beyer and colleagues. 105 Thirteen studies compared brachytherapy versus EBRT,105,119,126,135,136,170–172,182,189,205–207 one brachytherapy versus cryotherapy,203 nine brachytherapy versus RP,101,108–110,113,121,123,145,149 one AS versus EBRT versus RP,198 one brachytherapy versus cryotherapy versus RP,160 one brachytherapy versus cryotherapy versus EBRT versus RP,128 13 brachytherapy versus EBRT versus RP36,100,117,130,131,144,151,153,156,163,176,184,186 and one brachytherapy versus cryotherapy versus HIFU versus PDT. 103 Forty-four case series,52,98,99,102,104,106,107,111,114,116,120,122,124,127,129,132–134,138–143,146,150,154,155,158,159,161,162,166,173,174,181,185,187,188,191,195,199,202,204 including 20 prospective52,98,99,104,111,114,124,134,139–141,143,146,150,155,161,181,187,199,202 and 13 retrospective studies,102,106,107,127,132,133,138,154,162,166,173,185,188 were included in the primary review. Fourteen studies were case series of cryotherapy,52,102,114,122,124,129,138,139,154,158,166,188,202,204 20 of HIFU,98,99,106,107,116,120,127,132,133,142,143,150,159,161,162,173,174,185,191,195 one of laser therapy155 and nine of AS. 104,111,134,140,141,146,181,187,199 Table 4 summarises the number and types of included studies.

To further summarise the network of studies in the primary review, Table 5 is a matrix of the number of studies in the primary review by comparison/intervention.

Two studies were considered to include potential patient overlap: Ganzer and colleagues133 derived data from the multicentre-based @-Registry for 804 participants who were recruited between February 1993 and July 2009 and treated with HIFU in Lyon (France), Regensburg (Germany), Como (Italy) and Montpellier (France), while Blana and colleagues107 reported 356 HIFU participants recruited between February 1993 and October 2010 from the same registry, from nine centres. These studies were treated separately because Blana and colleagues, in addition to similar inclusion criteria reported by Ganzer and colleagues, also selected participants with anteroposterior prostate height of ≤ 24 mm and a treated volume of > 120% of the prostate volume, while Ganzer and colleagues also selected participants with a minimum follow-up of 3 years. Similarly, Uchida and colleagues191 reported the results of 72 consecutive participants treated with HIFU in different centres within an unspecified period of time. The same authors also reported data related to 517 participants recruited between January 1999 and December 2007 from a single clinical centre. 195 These data sets were treated as two separate studies because if any patient overlap existed it was likely to have a minor impact on meta-analyses, as the study sample sizes were significantly different.

Malcolm and colleagues160 and Hubosky and colleagues52 were, respectively, a NRCS and a case series conducted in the same centre around the same time period. The same number of participants enrolled by Hubosky and colleagues52 was enrolled into the cryotherapy arm of the study by Malcolm and colleagues. 160 However, the baseline characteristics of the patient groups were different and therefore the data sets were considered as separate studies.

The RCTs of brachytherapy versus RP by Giberti and colleagues49 and Crook and colleagues121 were conducted in Italy and Canada respectively; Donnelly and colleagues125 conducted a study of cryotherapy versus EBRT in Canada, and Paulson and colleagues168 conducted a study of EBRT versus RP in the USA. The NRCS of brachytherapy versus cryotherapy conducted by Williams and colleagues203 was set in the USA. About two-thirds of the NRCSs of brachytherapy versus EBRT were set in the USA, while one172 was conducted in Germany, two in Canada171,189 and two in both the Netherlands and Austria. 135,136 Of nine non-randomised studies of brachytherapy versus RP, there were three each from the USA101,110,123 and Germany,108,109,145 and one each from France,113 Canada121 and Japan. 149 The non-randomised study of AS versus EBRT versus RP198 was conducted in the Netherlands. Both non-randomised studies of brachytherapy versus cryotherapy versus RP were from the USA. 128,160 Of 13 studies of brachytherapy versus EBRT versus RP, 11 were conducted in the USA36,100,117,131,144,151,153,156,163,176,186 and one each in Spain and Australia. 130,184

Characteristics of study participants

A total of 72,259 study participants from 88 studies were enrolled; 26,129 had brachytherapy, 3995 had cryotherapy, 4000 had HIFU, 12 had laser therapy, 23 had PDT, 12,547 had EBRT, 19,961 had RP and 5592 had AS. Of these, 70,804 (99%), including 25,805 brachytherapy, 3964 cryotherapy, 3997 HIFU, 12 laser therapy, 23 PDT, 5437 AS, 12,426 EBRT and 19,140 RP participants, were included in the analyses. Table 6 shows the demographic and disease characteristics of the study participants.

Most studies reported either the mean or median age; 12 studies did not report this information. 36,100,103,128,133,150,163,168,182,203,205,207 The average age was similar across interventions.

At least half of the participants in all interventions were clinical stage T1, except in cryotherapy, where T1 participants constituted one-fifth. T2 participants made up about one- to two-fifths across all intervention groups.

About 20–25% of brachytherapy, cryotherapy, EBRT, RP and HIFU participants were Gleason 6, as were the majority of AS, laser therapy and PDT participants. The proportion of participants with a Gleason score of 7 ranged from 2.1% in AS to 22.1% in cryotherapy.

The average PSA ranged from 5.55 ng/ml in AS to 8.43 ng/ml in cryotherapy. Of those reporting PSA, 18 studies did not report it as a mean or median. 36,102,103,105,108,117,123,126,129,138–140,146,187,202,203,205,206

Thirty studies reported prostate size;49,99,106,107,111,113,116,120,122,123,125–127,130,132,133,139,142,143,155,159,161,162,172–174,185,188,191,195 most studies on HIFU and laser therapy and almost half the studies on AS reported the prostate size, whereas most of the studies on other interventions did not. The average prostate size reported ranged from 26.5 ml in HIFU to 45.0 ml in RP.

Categorising studies into low-, medium- and high-risk localised disease

As the results in Table 6 illustrate, the variety of differences in reporting of clinical stage, Gleason score and PSA made it impossible to categorise the studies according to the criteria described in Table 3. Although the inability to categorise studies according to the risk strata had no significant effect on comparisons including EBRT and RP, the inability to identify studies of people with low-risk localised disease meant that no comparison with AS would have been possible. After discussion with the expert advisory group, a pragmatic decision was made to categorise as studies of low-risk localised disease all those in which the Gleason scores of two-thirds of the patients were Gleason 6 or less in the ablative studies.

Thirty-four studies36,49,101,103,105,107,109,110,116,117,119,123,126,129,135,138,151,153,155,156,160,162,170–172,174,184,186,189,198,202,205–207 met this criterion and were compared with the AS participants in a subgroup analysis. This subset of participants included 9069 brachytherapy, 1377 cryotherapy, 5628 EBRT, 994 HIFU, 12 laser therapy, 23 PDT and 7840 RP participants at enrolment.

Focal ablative therapies

Each included ablative study was categorised depending on whether or not a focal approach was the primary intervention.

Overview of type of outcomes reported

Risk-of-bias/quality assessment

Forty-three studies, comprising 39 NRCSs36,100,101,103,105,108–110,113,117,119,121,123,126,130,131,135,136,144,145,149,151,153,156,160,163,170–172,176,182,184,186,189,198,203,205–207 and 4 RCTs,49,121,125,168 were assessed for risk of bias for the primary outcomes of this review using the Cochrane risk-of-bias tool. 89 Forty-four case series were assessed for methodological quality using the Review Body for Interventional Procedures (ReBIP) checklist. 52,98,99,102,104,106,107,111,114,116,120,122,124,127,129,132–134,138–143,146,150,154,155,158,159,161,162,166,173,174,181,185,187,188,191,195,199,202,204 One study which reported exclusively adverse events, but not other relevant outcomes, was not assessed for risk of bias. 128

Randomised controlled studies

The results of the risk-of-bias assessments for individual studies are shown in Appendix 9. The assessments are summarised in Figures 4–8.

FIGURE 4.

Summary of risk-of-bias assessments for RCTs reporting efficacy (n = 3).

FIGURE 5.

Summary of risk-of-bias assessments for RCTs reporting urinary function (n = 2).

FIGURE 6.

Summary of risk-of-bias assessments for RCTs reporting sexual function (n = 2).

FIGURE 7.

Summary of risk-of-bias assessments for RCTs reporting bowel function (n = 1).

FIGURE 8.

Summary of risk-of-bias assessments for RCTs reporting quality of life (n = 2).

Efficacy

Three studies were assessed for risk of bias of efficacy outcomes. 49,125,168 Of these, only the study by Giberti and colleagues49 was considered to be at low risk for sequence generation and others were unclear. None provided adequate information for the assessment of allocation concealment.

Urinary function

Two studies were assessed for risk of bias of urinary function outcomes. 49,121 The study by Giberti and colleagues49 was considered low risk of bias for sequence generation whereas this was unclear in the study by Crook and colleagues,121 and both were judged as unclear risk of bias for allocation concealment.

Sexual function

Two studies were assessed for risk of bias of sexual function outcomes. 49,121 The study by Giberti and colleagues49 was considered to be at low risk for sequence generation, whereas there was insufficient information to assess sequence generation in that by Crook and colleagues121 or allocation concealment in either study. 49,121

Bowel function

Only one study49 was assessed for risk of bias of bowel function outcomes; it was considered to be at low risk of bias for sequence generation and unclear for allocation concealment.

Quality of life

Two studies were assessed for risk of bias of quality of life outcomes. 49,121 The study by Giberti and colleagues49 was considered to be at low risk for sequence generation whereas that by Crook and colleagues121 was unclear. Both studies were judged as unclear for allocation concealment. 49,121

Non-randomised controlled studies

The results of the risk-of-bias assessments for individual studies are shown in Appendix 9. The assessments are summarised in Figures 9–13.

FIGURE 9.

Summary of risk-of-bias assessments for NRCSs reporting efficacy (n = 21).

FIGURE 10.

Summary of risk-of-bias assessments for NRCSs reporting urinary function (n = 23).

FIGURE 11.

Summary of risk-of-bias assessments for NRCSs reporting sexual function (n = 21).

FIGURE 12.

Summary of risk-of-bias assessments for NRCSs reporting bowel function (n = 16).

FIGURE 13.

Summary of risk-of-bias assessments for NRCSs reporting quality of life (n = 13).

Efficacy

Twenty-one studies were assessed for risk of bias of efficacy outcomes. 36,101,103,105,109,119,123,126,135,136,144,149,151,170,171,182,184,189,205–207 Of these, nine103,105,109,119,135,144,184,205,206 were considered to be at low risk of bias for confounding and one182 was unclear.

Urinary function

Twenty-three studies were assessed for risk of bias of urinary function outcomes. 103,108–110,113,117,121,126,130,131,145,149,153,160,163,172,176,182,184,186,189,203,206 Of these, 11109,121,130,149,153,172,176,184,186,203,206 were considered to be at low risk of bias for confounding. The studies by Frank and colleagues131 and Shah and colleagues182 were unclear.

Sexual function

Twenty-one studies were assessed for risk of bias of sexual function outcomes. 100,103,109,110,113,117,121,126,130,131,149,160,163,172,182,186,189,198,203,206,207 Of these, seven109,130,172,184,186,189,203 were considered to be at low risk of bias for confounding. The studies by Barret and colleagues103 and Frank and colleagues131 were unclear.

Bowel function

Sixteen studies were assessed for risk of bias of bowel function outcomes. 109,110,113,117,126,130,131,149,156,160,172,184,186,189,203,206 Of these, seven109,130,156,184,186,189,206 were considered to be at low risk of bias for confounding. The studies by Williams and colleagues203 and Frank and colleagues131 were unclear.

Quality of life

Thirteen studies were assessed for risk of bias of quality of life outcomes. 109,110,121,130,131,145,149,153,160,172,176,184,198 Of these, seven were considered to be at low risk of bias for confounding. 109,121,131,149,153,172,184 The study by Reeve176 was unclear.

Case series

The ReBIP checklist was used to assess the methodological quality of the case series. Studies with all items scored as ‘no’ or ‘unclear’ were considered at high risk of bias. All case series included in this review were judged as having a high risk of bias. The results of the quality assessment are summarised in Figure 14 and further details are provided in Appendix 9.

FIGURE 14.

Summary of quality assessments of the case series in the primary review.

Number and types of studies included

All included studies were case series; six were studies of salvage RP,209–211,213–215 two of salvage cryotherapy208,212 and one of salvage HIFU. 120 Data were collected prospectively in three of the included studies212,213,215 and retrospectively in a further three,120,208,211 and in the remaining studies patient enrolment was unclear. Table 7 summarises the number and types of included studies.

The study by Chin and colleagues208 and that by Robinson and colleagues212 were conducted in Canada; those by Gheiler and colleagues,210 Neerhut and colleagues211 and Tefilli and colleagues214 were conducted in the USA; and those by Darras and colleagues,209 Seabra and colleagues213 and van der Poel and colleagues215 were conducted in Belgium, Brazil and the Netherlands respectively.

Characteristics of study participants

A total of 400 participants were enrolled in nine studies; 164 had salvage cryotherapy, 71 had salvage HIFU and 165 had salvage RP. Three hundred and eighty-eight (388) (97%) participants, encompassing 164 salvage cryotherapy, 71 salvage HIFU and 153 salvage RP patients, were included in the final outcome analyses. Table 8 summarises the baseline characteristics of the study participants.

The mean age of salvage cryotherapy participants was comparable with that of salvage RP participants. All interventions were comparable in terms of participants with clinical stage T2 or less. The Gleason scores of enrolled participants were not comparable, as the proportion of participants with Gleason scores of 6 or less who underwent salvage RP was double that of those who underwent salvage cryotherapy, and vice versa for participants with Gleason scores of 8 or more. There was no information on Gleason score for participants who underwent salvage HIFU. It was not possible to comment on the PSA and prostate size because data were limited.

Overview of studies reporting the main outcomes of the review

Quality assessment

All case series included in the salvage review were judged at high risk of bias as all methodological items were scored as ‘no’ or ‘unclear’ on the ReBIP checklist. The results of the quality assessment are summarised in Figure 15 and further details are provided in Appendix 9.

FIGURE 15.

Summary of quality assessments of the case series in the salvage review.

Cancer-related efficacy outcomes

Adverse effects

Procedural complications

Data on short-term adverse events related to the use of cryotherapy, including dysuria, urinary retention, urethral sloughing, infection, stricture, bladder neck contracture, bladder spasm, rectal pain/bleeding and fistula, are presented below. Abstracted data concerning other specific adverse events not included below are detailed in Appendix 10. Given the variety of definitions and periods of follow-up between studies, a degree of caution should be used in interpretation of these results.

Quality of life

Only one case series of people having cryotherapy reported on quality of life outcomes. 124 The data were insufficient to enable us to assess any difference in this outcome compared with either EBRT or RP.

Further prostate cancer treatment

The need for reintervention within 2 years of initial procedure using further cryotherapy was reported by six studies of people undergoing primary cryotherapy. 114,122,125,129,154,204 The rates of reintervention ranged from 1% to 15% with a median rate of 9% across the studies.

Within 6 months of initial treatment, Donnelly and colleagues125 reported that 11% of people treated with cryotherapy received hormonal androgen deprivation therapy and 3% were placed in a watchful waiting programme. In contrast, Caso and colleagues114 reported the rate of further cancer treatment using any modality at a median follow-up of 2 years to be 12%.

Focal cryotherapy

Of the 19 studies describing the results of primary cryotherapy,52,102,103,114,122,124,125,128,129,138,139,154,158,160,188,202–204,216 six used a focal ablative approach. 103,129,138,166,188,202 Given the low number of studies reporting on the use of focal cryotherapy and the diversity of outcomes reported in each study, no formal subgroup meta-analyses could be undertaken for most of the outcomes, and therefore a descriptive summary of reported findings in relation to the overall comparative meta-analysis is given below.

Use of cryotherapy versus active surveillance for people with low-risk prostate cancers

As described in the methods of the systematic review (see Chapter 3), any comparison with AS necessitated that the included studies contained low-risk patients only. Five studies reporting the outcome of cryotherapy met the low-risk disease criteria (described in Chapter 3) for inclusion. 103,129,138,160,202 The studies variably reported comparative outcomes of overall survival, functional outcomes (urinary incontinence and erectile dysfunction), quality of life and need for further cancer treatment.

Cancer-related efficacy outcomes

Adverse effects

Procedural complications

Data on short-term adverse events related to the use of HIFU, including dysuria, urinary retention, urethral sloughing, infection, stricture, bladder neck contracture, bladder spasm, rectal pain/bleeding and fistula, are presented below. Abstracted data concerning other specific adverse events not included below are detailed in Appendix 10. Given the variety of definitions and periods of follow-up between studies, a degree of caution should be used in interpretation of these results.

Quality of life

Two case series of people undergoing HIFU reported on a variety of quality of life outcomes,98,195 but none of the measures were the same between studies. The data were, therefore, insufficient to inform on any difference in quality of life following HIFU compared with either EBRT or RP (see Appendix 10, Table 88, for full details).

Further prostate cancer treatment

The need for reintervention using further HIFU within 2 years of initial procedure was reported in three studies of people undergoing HIFU. 116,161,173 The rates of reintervention were 3%,173 12%161 and 31%116 respectively.

Within 6 months of initial treatment, Pinthus and colleagues173 reported that 1% of patients treated with HIFU received hormonal androgen deprivation therapy and 7% were placed in an AS programme; 1.5% received RP and 1% EBRT. At 4 years, Misrai and colleagues162 reported that 12% received EBRT, 6% received hormonal androgen deprivation therapy and 1% received RP. In contrast, at 8 years, Sumitomo and colleagues185 reported that 2% received EBRT, 22% received hormonal androgen deprivation therapy and 2% received RP.

Focal high-intensity focused ultrasound

Of the 21 studies reporting outcomes in people receiving HIFU,98,99,103,106,107,116,120,127,132,133,142,143,150,159,161,162,173,174,185,191,195 four used a focal HIFU approach. 98,99,103,127 Given the low number of studies reporting on the use of focal HIFU and the diversity of outcomes reported in each study, no formal subgroup meta-analyses could be undertaken, and therefore a descriptive summary of reported findings in relation to the overall comparative meta-analysis is given.

Use of high-intensity focused ultrasound versus active surveillance for people with low-risk prostate cancer

As described in the methods of the systematic review (see Chapter 3), any comparison with AS necessitated that the included studies contained low-risk patients only. Two studies of people following HIFU met the low-risk patient criterion for inclusion. 116,162 The studies variably reported comparative outcomes of overall survival, functional outcomes (urinary incontinence and erectile dysfunction), quality of life and need for further cancer treatment.

Cancer-related efficacy outcomes

Adverse effects

Procedural complications

Data on short-term adverse events related to the use of brachytherapy, including dysuria, urinary retention, infection, stricture, bladder neck contracture, rectal pain/bleeding, fistula and toxicity, are presented below. Abstracted data concerning other specific adverse events not included below are detailed in Appendix 10. Given the variety of definitions and periods of follow-up between studies, a degree of caution should be used in interpretation of these results.

Quality of life

Quality of life was not reported consistently enough across studies to perform a meta-analysis. The most robust evidence came from a single RCT of brachytherapy versus RP. 49 The patients in the study reported similar significant decreases in some functional and symptom European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire – Core 30 (EORTC-QLQ-C30) scales after 6 months and 1 year regardless of the treatment, and both groups of patients reported a normal global health after 1 and 5 years.

A similar pattern at 2-year follow-up was observed in the non-randomised study130 that compared people receiving brachytherapy, RP and EBRT. Health-related quality of life (HRQoL) initially decreased across all the treatment modalities, and made a partial recovery by 2 years (see Appendix 10, Table 88 for full details).

Further prostate cancer treatment

Within 3 months of initial treatment, Giberti and colleagues49 reported that 2.5% of people treated with brachytherapy received hormonal androgen deprivation therapy, 2.5% received EBRT and 3% received RP. In contrast, Pickles and colleagues171 reported a rate of hormonal androgen deprivation therapy at a follow-up of 5 years to be 5%, compared with 8% for people who initially received EBRT.

Focal brachytherapy

Of the 39 studies on brachytherapy, only Barret and colleagues103 reported a focal technique and included 12 participants at enrolment and in the final outcome analyses. Given the low number of studies (and people) reporting on the use of focal brachytherapy, no further data exploration was undertaken.

Use of brachytherapy versus active surveillance for people with low-risk prostate cancers

As described in the methods of the systematic review (see Chapter 3), any comparison with AS necessitated that the included studies contained low-risk patients only. Twenty-four studies reporting the outcome of brachytherapy met the low-risk disease criterion for inclusion. 36,49,101,105,109,110,117,119,123,126,135,151,153,156,160,170–172,184,186,189,205–207 The studies variably reported comparative outcomes of overall survival, functional outcomes (urinary incontinence and erectile dysfunction), quality of life and need for further cancer treatment.

Cancer-related efficacy outcomes

Functional outcomes

For focal laser ablative therapy,155 there was no significant change in urinary or erectile function at 3 and 6 months postoperatively compared with the preoperative status.

For PDT,103 at a median follow-up of 9 months, there was no difference in urinary function nor erectile function between PDT and the other comparators (HIFU, cryotherapy and focal brachytherapy). However, for the intragroup comparison of patients who underwent PDT, erectile dysfunction appeared to be worse after treatment than prior to treatment (a difference in median scores of 10 on the IIEF-5 score); the authors did not specify whether or not this result was statistically significant. There was no difference in urinary function between the pre- and postoperative states in patients who underwent PDT. In terms of continence, all patients were reportedly continent postoperatively.

Adverse events

For focal laser ablative therapy,155 there were no perioperative complications. Postoperative morbidity was minimal and self-limiting; this included perineal discomfort (25% of patients), mild haematuria (16.7% of patients) and haematospermia (16.7% of patients).

For PDT,103 the results for treatment-related complications were not reported separately. However, all of the reported complications involved patients who underwent either HIFU or cryotherapy; hence it is assumed that no patients who underwent PDT had any complications.

Cancer-related efficacy outcomes

Only two studies on salvage ablative therapies reported on cancer-related outcomes,208,212 and both studies involved salvage cryotherapy. In contrast, all six studies on salvage RP209–211,213–215 reported on cancer-related outcomes. The data were limited by heterogeneity of outcome definition, different time points of outcome measurement and different means of reporting (e.g. biochemical control vs. failure).

Functional outcomes

Only three studies on salvage ablative therapies reported on functional outcomes: two studies on salvage cryotherapy179,208 and one on salvage HIFU. 120 Six studies of salvage RP reported on functional outcomes. 209–211,213–215 The data were limited by heterogeneity of outcome definition, different time points of outcome measurement and different means of reporting (e.g. urinary continence vs. incontinence).

Quality of life

Only one study on salvage ablative therapy reported on quality of life outcomes; this was Robinson 2006212 on salvage cryotherapy. One study on salvage RP reported on quality of life outcomes. 214 The data were limited by heterogeneity of the different quality of life measures used, different time points of outcome measurement and different means of reporting (e.g. total score vs. individual component score).

Adverse events

Only three studies on salvage ablative therapies reported on adverse events: two studies on salvage cryotherapy208,212 and one on salvage HIFU. 120 Six studies of salvage RP reported on adverse events. 209–211,213–215 For salvage cryotherapy, at a median follow-up of 18.6 months, the incidence of adverse events was relatively low, ranging from 2% (bladder neck stenosis) to 3% (rectourethral fistula). The corresponding figure for salvage RP, within a similar period of follow-up, was 4.8–6% (rectovesical fistula) and 3–25% (bladder neck stenosis or anastomotic stricture). For salvage HIFU, at 15 months follow-up, the incidence of rectourethral fistula was 6% and that of bladder neck stenosis was 17%. The data were limited by the relatively low number of patients and low event rates.

Assumptions

Most of the assumptions inherent in the modelling process were derived from definitions and estimations of the driving parameters (see Estimation of probabilities used within the model). The importance of these assumptions in determining model output was estimated with elasticity analysis, as described in Estimation of utilities used within the model.

Process overview and scheduling

The structure of the simulation model is described in the care pathways constructed for all ablative and comparative therapies for prostate cancer (Figure 16 in general and Figures 17–20). This care pathway forms the basis of a conceptual model of the processes to be simulated, and consists of three different elements: states, events and the transitions between them. A state is a stage in the model in which the patient spends at least one time step. Events are stages in the network that take up less than one time step. Each individual could therefore undergo one or more events plus a single state in each time step of the model. Transitions are the probability that an individual passes between different states and events. The care pathway does not claim to capture every possible patient trajectory (that is, a single route through the care pathway), as factors connected to the patient and their health-care team may result in treatment decisions that are unique to their individual circumstances. However, this care pathway was scrutinised by our expert advisory group and is considered to be definitive for 95% or more of all possible patient trajectories.

FIGURE 17.

Model pathways for ablative therapies. MORT, mortality.

FIGURE 18.

Care pathway for brachytherapy. HT, hormonal therapy; MORT, mortality.

FIGURE 19.

Care pathway for EBRT. HT, hormonal therapy; MORT, mortality.

FIGURE 20.

Care pathway for RP. HT, hormonal therapy; MORT, mortality.

The conceptual model can be conceived as a network or graph. In mathematical terms, a network is expressed as a set of vertices and a set of edges which connect them. In this context, the vertices were states and events, and the edges were the transitions. This mathematical framework can be described in terms of an adjacency matrix for this network, and this adjacency matrix served as the transition matrix to determine the next event or state experienced in each modelled time step.

Beginning at the initial event of diagnosis, the state of each simulated patient at each time step was determined by the transition matrix. Within a modelled time step, a patient could also experience one or more events. Over time the patient could receive different types of therapy, experience different adverse events or spend time in one of three surveillance states: AS which occurs before primary therapy, surveillance which occurs after primary therapy and before biochemical failure, and follow-up surveillance which occurs following biochemical failure and salvage treatment. The simulation ended once all patients had entered one of the three ‘sink’ states: operative mortality, non-cancer mortality and cancer mortality.

The model employed a 6-month time step. All driving variables were probabilities that were usually calculated as yearly probabilities (P_12m); these were scaled to 6-month probabilities (P_6m) thus:

Design concepts

The model was a modified Markov chain simulation model, where subsequent status was determined by current status multiplied by a matrix of transition probabilities. The basic design of a Markov chain model was altered to permit the state variables (age and severity) to have an impact on driving variables (transition probabilities).

Stochasticity

Stochasticity was incorporated into the model by simulating probabilities of changes in health status by sampling random deviates from a uniform distribution in the range of 0–1. Subsequent health status for each simulated patient trajectory was determined by the value of the random deviate compared with the cumulative probability of all destination states and events from the current health status. For states, remaining in the current state and entering the non-cancer-related mortality sink state were always possible destinations; neither of these options was available for events.

Observation

In each time step, the state of each simulated patient was recorded along with all events experienced during the time step. Each state and event was assigned a monetary cost. Utilities assigned to each state and event were converted into QALYs by calculating the product of all utilities experienced in each modelled year and then summing over the survival of the patient (i.e. the time before entering one of the three sink states).

Initialisation

The model was initialised with 1000 patients. Age of each patient was randomly sampled from a Poisson distribution with λ = 70 years. Severity for each patient was determined from data in the review of clinical effectiveness reported earlier [see Chapter 3, Primary review (quantity and quality of included studies)].

Inputs

There were two inputs to the model. The first input was an edge list of possible transitions in the model, with associated driving variables (transition probabilities). The second input was a vertex list of the monetary cost and the utility associated with each event or state. The data used for transitions, cost and utilities are described below (see Estimation of probabilities used within the model, Estimation of costs used within the model and Estimation of utilities used within the model).

Submodel: side effects

The three side effects of urinary incontinence (UI), erectile dysfunction (ED) and bowel dysfunction (BD) were each simulated independently with submodels. The side effects submodel was a Markov chain model dependent on the current state or event of the individual, or the current side effect status of the individual. Patients undergoing primary treatment had a probability of developing a dysfunction at an initial prevalence A, and therefore had a probability of maintaining function of 1 − prevalence A. In subsequent time steps, development of a dysfunction or recovery to functionality was determined by a second prevalence, B (Table 44). Patients in AS did not develop dysfunctions until after active treatment commenced (where their state is changed to primary treatment by the main model), and so were unable to develop or recover from dysfunctions prior to this. Patients in one of the three deceased states also remained in their current state (unable to develop or recover from dysfunctions).

The two prevalences, A and B, allowed for an initial post-treatment prevalence of a side effect that was different from the long-term prevalence of the side effect; typically, prevalence A was higher than prevalence B, but this was not always the case.

Age distribution of cohort being modelled and the probability of death from causes other than prostate cancer

The age distribution of the cohort was randomly sampled from a Poisson distribution with a λ value of 70.

At any point in the model, there is a risk of death from all causes, including prostate cancer. The interventions under investigation might alter the prostate cancer-specific component of this mortality but would not be expected to affect other causes of mortality. Non-cancer mortality was calculated from age- and sex-specific UK life expectancy and mortality tables for the UK produced by the Office for National Statistics (ONS). 219,220 These data were resolved into a lognormal equation of male age-specific mortality rates using generalised linear modelling. This equation explained 97% of the variation in the published data.

Selection of primary treatment or active surveillance on diagnosis

Five separate model runs were conducted using the entire cohort of 1000 simulated patients, one model run for each of the main treatments (HIFU, cryotherapy, brachytherapy, EBRT and RP). Separate model runs meant that it was not necessary to use routine data sources to estimate the relative proportions of patients diagnosed with localised disease receiving each treatment as their primary therapy.

However, routine data sources were required to derive an estimate for parameter values regarding the proportion of patients receiving adjuvant treatments (e.g. EBRT or hormone therapy), and the proportion of people who received AS prior to any primary therapy.

Data from the National Cancer Intelligence Network provide information on treatments given to 18,839 diagnosed patients in 2009 (for a further 16,008 the initial treatment strategy was reported as unknown). 221 These data were not available separately for patients in different stages of disease at diagnosis. However, the British Association of Urological Surgeons (BAUS) Cancer Registry has published information on the initial treatment strategies used for patients diagnosed in 2007, and these data are available separately for various PSA level thresholds (0–5, 6–10, 11–15, 16–20, 21–50, 50 +). 222 We used a threshold PSA of ≤ 20 as a proxy for localised disease and used both sources to estimate the required parameter values.

In 2009, approximately 5104 patients received AS as their primary treatment strategy following diagnosis. 221 This equates to 27.1% of the 18,839 patients for whom primary treatment was known.

The probability of staying in AS rather than moving to a primary therapy was determined from data in the review of clinical effectiveness described in Chapters 4–6 (Table 45).

Probabilities related to primary treatment

Perioperative adverse events

The systematic review of clinical effectiveness identified studies that had reported information on perioperative adverse events. Two clinicians (RP and TL) graded each of these adverse events based on expected severity and subsequent management. This rating system informed the Clavien–Dindo rating approach. 80 An average probability for the occurrence of a perioperative adverse event for each grade was calculated using the reported information from the review. The model accounted for differences between the treatments in terms of perioperative adverse events by costing additional days in hospital caused. Information on additional length of stay in hospital for different Clavien–Dindo ratings was taken from a study by Prentis and colleagues,224 whereby ratings of < 3 and ≥ 3 resulted in 4 and 15 additional days’ stay respectively (Table 49).

Biochemical recurrence after primary treatment

Salvage treatment for localised recurrence

Treatment for systemic recurrence after primary or salvage treatment

It was assumed that on progression following salvage treatment, or on systemic recurrence following primary treatment, patients would be treated for advanced disease, with ‘watchful waiting’ being conducted initially and alternative treatment options (which in clinical practice will depend on individual patient circumstances) being hormonal therapy and castrate-resistant stage therapies (including chemotherapy) as well as palliative treatment. Palliative treatment might also involve drug treatment, but to treat sequelae of the prostate cancer (e.g. bone metastases) rather than the prostate cancer itself.

Table 58 shows the yearly transition probabilities for those who move from hormonal therapy, castrate-resistant disease and other palliative treatments to subsequent care and events.

We used guidelines on the treatment of the disease at this stage to estimate the typical processes of care at this stage in the care pathway. We assumed that patients at this point would initially receive hormonal therapy, with the rate of progression to hormone refractory disease taken from available guideline evidence on treatment for metastatic disease. 73 The initial probability of response was estimated at 80%, with 60% still showing progression-free response at 3 years. 73

Overall survival has been separately estimated at a median of 5 years,226 with 7% surviving beyond 10 years. 225 Similarly, further review evidence on survival following chemotherapy treatment for advanced cancer was used to inform the model for castrate-resistant disease. On progression with hormone therapy, the probability of staying in the castrate-resistant stage for ≥ 1 time step was considered equivalent to the yearly probability of response to chemotherapy (0.52). 226 Palliative treatment was considered to confer a 6-month survival on average (though cancer mortality could also occur prior to having any palliative treatment). 227

Probability of longer-term adverse events (used in submodels of adverse events)

Values for the two prevalences (A and B) were calculated from data in the systematic review. For each of the three adverse events it was assumed that within the first 6 months the rate would differ from any longer-term trend. Prevalence A was calculated as the median for all sources reporting the prevalence of the adverse event at a follow-up time of ≤ 6 months. It was assumed that after 6 months, the prevalence would settle to a constant rate. All data on each adverse event that were reported for a follow-up time of beyond 6 months were converted to a yearly rate and then the average was taken to calculate prevalence B. The results are summarised in Table 59.

Treatment costs associated with primary treatments

Table 60 shows the cost estimates used in the model for AS, Table 61 the cost estimates for primary treatment costs and Table 62 the cost estimates for the follow-up surveillance state.

Treatment costs associated with biochemical recurrence after primary treatment

Based on elevated PSA levels observed while under surveillance state, biochemical recurrence can entail two different types of diagnosis event: local recurrence and metastatic recurrence. Table 63 shows the costs of diagnosing local and metastatic recurrence.

Summary of costs used in the model

Costs used in the model are all summarised in Table 64.

Costs and utilities used in the submodel of adverse events

Time in each state of dysfunction for all three side effects incurred a cost which was added to the yearly costs to obtain lifetime totals for each patient. Costs used are listed in Table 65.

Diagnosis events

Where multiple sources of utility values for particular parameters were available, median values from the literature reviewed were used, which were then calibrated to the EQ-5D. For local recurrence, utility values were estimated on the basis of values taken from four studies. 243,247–249 For systemic recurrence, utility values were estimated on values taken from two studies. 247,250

Primary treatments

For many primary treatment events, such as brachytherapy, cryotherapy, etc., there were no utility data available. It was assumed that the utility values for these treatment events were the same as that used for surveillance. This seemed a reasonable assumption as this was estimated to be the same as the utility value for EBRT.

Where EQ-5D scores were available from one source for multiple time points (as with EBRT),239 the percentage improvement from baseline to 6 months post intervention was calculated. This was then calibrated by the EQ-5D value of initial diagnosis. It was assumed that the utility values for adjuvant EBRT with and without hormone therapy were the same as this. The utility value for RP with lymphadenectomy was assumed to be the same as that for RP alone owing to the absence of data.

Further cancer treatment

There were no utility data for any of the salvage treatments included in the model. We therefore estimated this (for all salvage treatments), taking the average of utility values for local recurrence, 6 weeks post RP and surveillance at 12 weeks.

Summary of costs and utilities used in the submodel of adverse events

Time in each state of dysfunction for all three side effects incurred a cost and utility which were (respectively) added to/multiplied by the yearly costs and utilities to obtain lifetime totals for each patient. Costs and utilities used are listed in Tables 65 and 67.

Economic outputs

The economic outputs of the model included:

• Total costs per patient over the patient lifetime. These data tended to be highly skewed as some patients survived in the model for a long time but also experienced a number of very high-cost events. These were then summarised at a population level to produce average total cost over the patient lifetime for each initial treatment.

• Total QALYs of each patient. As noted earlier, this was calculated by summing the yearly products of the within-year utilities for each state and event. QALYs also tended to have a highly skewed distribution as some patients experienced an early death or experienced events that greatly reduced the amount of QALYs they could gain. These were then summarised at a population level to produce average QALYs for each initial treatment.

• Incremental mean costs.

• Incremental mean QALYs.

• Incremental cost per QALY gained.

• Net monetary benefits.

Within the base-case analysis, we adopted a NHS perspective and discounted costs and QALYs at the recommended 3.5% discount rate. 231 All costs and QALYs are for a lifetime time horizon and all monetary values are expressed in 2011–12 prices.

The base-case analysis has assumed that biochemical recurrence does not differ across the procedures, which is consistent with the evidence in the review of effectiveness presented in Chapters 4–7. However, an alternative analysis where biochemical recurrence varies according to the results of the meta-analysis of clinical effectiveness is also reported.

Sensitivity analysis

We addressed uncertainty by conducting probabilistic sensitivity analyses and deterministic (e.g. one-way) sensitivity analyses. The probabilistic sensitivity analysis involved running 1000 iterations of the model for each intervention considered for each analysis. These data were then used to prepare cost-effectiveness plots and cost-effectiveness acceptability curves. 258 These curves provide an estimate of the likelihood that an intervention would be considered cost-effective at different threshold values for society’s willingness to pay for either a recurrence avoided or a day at usual activities.

The following deterministic sensitivity analyses were considered. A new intervention, AS, was introduced as an alternative to initial active therapy for localised prostate disease. In effect, this is a policy of delayed and selective treatment, which might be a viable option where disease is unlikely to become symptomatic over the expected lifetime of the individual or where the expected harms (in terms of side effects) would be worse than any symptoms currently experienced. This analysis was facilitated because it was assumed in the base-case analysis that, for each active treatment, a period of AS would take place for approximately 20% of people. In the modelling, this means that approximately 20% of the model runs for each active treatment involved AS. These data have been used to construct an additional comparator, AS, to allow cost-effectiveness analysis to be explored.

Internal consistency checks

With respect to face validity the structure of the model and all data inputs were scrutinised by the research team and the external advisory group to ensure that the model structure suitably reflected the decision problem addressed and that data inputs and methods to assemble these inputs seemed plausible.

The elasticity analysis provided a further computational validity in that it explored the importance of model transitions. This provided a check on the mathematical logic of the model and allowed the modelling code to be tested for errors. Counterintuitive results became the focus of further investigation. Further, sensitivity analysis (not reported) was used to explore whether or not data had been incorporated correctly.

External validity

Alternative models comparing the cost-effectiveness of these interventions are not available and so this approach was not available to check external validity. However, the results of the model were checked with experts to assess their face validity. The model was also used to produce outputs that could be compared with data not used in the model (because it was not reported in sufficiently disaggregated form to be incorporated into the model).

Any issues with the internal or external validity of the model or its outputs were resolved prior to producing the final results reported in the next chapter.

Base-case analysis: equal risk of biochemical recurrence

When making the assumption that biochemical recurrence is equivalent, the choice between interventions is driven by three factors: (i) the cost of the interventions; (ii) perioperative complications; and (iii) the impact of long-term complications. Table 69 shows the incremental cost-effectiveness analysis for the comparison of the different interventions. These data are derived from the Monte Carlo simulations. As this table illustrates, HIFU is, on average, less costly per patient and results in more QALYs than any of the other interventions. However, this analysis is potentially misleading as it does not display the imprecision surrounding estimates of costs, QALYs and cost-effectiveness. This imprecision can be portrayed by plotting the costs and QALYs for each intervention (Figure 22) and, as was described in Chapter 9, these data can be displayed as cost-effectiveness acceptability curves (Figure 23).

FIGURE 22.

Base-case analysis: plots of costs and QALYs for each intervention.

FIGURE 23.

Base-case analysis: cost-effectiveness acceptability curves.

As Figure 22 illustrates, there is a wide variation in cost and QALYs for each intervention. For all interventions, the majority of individuals in the Monte Carlo simulation have relatively modest QALYs (< 5) but with considerably more variation in cost, which reflects the varying intensities of care that they receive over time. However, a small number of individuals for each intervention experience very low cumulative costs and considerably more QALYs, reflecting the possibility that some prostate cancers will not necessarily be problematic and might require considerably less care.

Figure 23 shows that should society not be willing to pay anything for an additional QALY, the intervention most likely to be cost-effective is EBRT, with an approximately 50% likelihood of being considered cost-effective. HIFU is more likely to be more costly than EBRT but provides more QALYs, hence as society’s willingness to pay for a QALY increases, the likelihood that HIFU would be considered cost-effective also increases. Thus, at threshold values for society’s willingness to pay for a QALY that might be considered worthwhile – for example, between £20,000 and £30,000 per QALY231 – there is a 70% likelihood that HIFU would be considered cost-effective. Over the same range, the other interventions (RP, cryotherapy and brachytherapy) have a very low likelihood of being considered cost-effective. It should be noted, however, that, as Figure 22 illustrates, the interventions are in fact similar and the results shown in Figure 23 are driven by small differences in costs and QALYs.

Alternative analysis using the results from the meta-analysis for biochemical recurrence

As an alternative to the base-case analysis, the results in this subsection assume that the results of the meta-analysis of biochemical recurrence are the most appropriate to use in the model. Table 70 shows that the rank ordering of interventions has now changed and that EBRT is now the least costly and least effective intervention, but with HIFU dominating the other treatments. In this analysis, HIFU is associated with an incremental cost per QALY that is beyond the threshold level generally considered acceptable for society. 231 However, these average data are very sensitive to the skewed data and do not illustrate the precision surrounding the estimates.

Figure 24 shows the plots of costs and QALYs for each intervention and these are broadly similar to the plots shown in Figure 22. Likewise, the cost-effectiveness acceptability curve for this analysis (Figure 25) shows a broadly similar pattern to that shown in Figure 23. Again, HIFU is most likely to be considered cost-effective at the threshold values for willingness to pay for a QALY that society might be willing to pay. However, the same caveats as noted above also apply.

FIGURE 24.

Plots of costs and QALYs for each intervention when risk of biochemical recurrence is based on the results of the meta-analysis.

FIGURE 25.

Cost-effectiveness acceptability curves when risk of biochemical recurrence is based on the results of the meta-analysis.

Primary ablative therapy

Salvage ablative therapy

Statement of principal findings

The first stage of the economic analysis was an elasticity analysis. The elasticity analysis helped identify which transition probabilities were the principal determinants of survival. This analysis was conducted for each intervention and showed that many of the principal determinants of survival were related to the characteristics of the initial cancer and many of the probabilities of outcomes following recurrence. Of moderate importance was the performance of individual therapies in preventing or delaying recurrence. In the economic evaluation that compared alternative interventions, the probabilities of events following recurrence were generally the same for all interventions and hence their effect on estimates of cost-effectiveness would be entirely caused by differences in recurrence rates between interventions.

With respect to the economic evaluation itself, the results of this analysis suggest that HIFU might be the intervention that is most likely to be considered cost-effective when assessed against threshold values for a cost per QALY that society might be willing to pay. 231 There is marked uncertainty within the analyses as plausible extremes would suggest that EBRT may also be most likely to be considered cost-effective in some circumstances. Over the limited range of analyses, cryotherapy, brachytherapy and RP were unlikely to be viewed as cost-effective over the threshold values for society’s willingness to pay for a QALY that were considered. It is, however, important to note that given the uncertainties surrounding parameter estimates and the similarities in costs and QALYs estimated (as illustrated by Figure 22), it is quite possible that plausible combinations of data inputs could be identified that could make these interventions appear cost-effective.

Thus, the results presented here are unlikely to be sufficient to form recommendations to change practice, but they do indicate that further robust studies around the relative effectiveness and cost-effectiveness of HIFU and EBRT as treatment options for localised prostate cancer may be most useful.

Clinical effectiveness

The main strength of the study is the systematic approach taken to review the literature. Exhaustive systematic searches were made of the major electronic databases. All potentially relevant studies were reviewed for eligibility. The risk of bias of included comparative studies and quality assessment of included case series were assessed using the best available tools. To prevent any biases caused by selective data abstraction, all outcomes were predetermined by both expert panel and patient focus group consensus. Any data were extracted using standard forms. Despite these efforts it is possible that some relevant data remained hidden as a result of non-publication.

In total, 121 reports were included. 36,49,52,98–215 Although this number of studies seems impressive, not every study contributed data to each outcome. Furthermore, differences in reporting between studies also limited the opportunities for robust meta-analysis. Given the limited evidence base, the CrIs around many of the estimates of differences were wide and included differences that would be clinically important but could favour any of the therapies under investigation. Another major limitation resulted from the majority of comparisons using case series, with few head-to-head comparisons of ablative therapies against current practice. The estimates were therefore generated using indirect comparisons. Like all analyses, they require assumptions to be made that may or may not be reasonable. In the context of this analysis, an important assumption is that the studies in each meta-analysis were representative of a similar population (i.e. the clinical and demographic characteristics of the people were similar across studies). Data in Table 2 demonstrated that the assumption had broad face validity, but that there were some differences, such as a slight increase in the average clinical stage for people in the EBRT study. Accordingly, the results should be interpreted with a large degree of caution.

A further methodological limitation that frustrated pooled analysis was the use of differing definitions and measures of functional outcomes for urinary, erectile and bowel dysfunction. The variety of different ways of measuring dysfunction reduced the ability to narratively compare data or to conduct a comprehensive meta-analysis. Although in part the difficulty is reflected by changing measurement methodology over time, it will remain a problem until consensus on important outcome measurements in this clinical area can be agreed. Initiatives such as the UK Medical Research Council- and European Union-funded Core Outcome Measures in Effectiveness Trials (COMET) or Consensus-based Standards for the Selection of Health Measurement Instruments (COSMIN) may be useful in this context. Such initiatives help patients, clinicians and researchers to develop a standardised set of outcomes that should be measured and reported as a minimum in all clinical trials of a specific condition, thereby making it easier to contrast and synthesise the results of trials.

Identifying outcomes that can be used to compare ablative therapies with both AS and RP or EBRT was challenging. Long-term survival is a key outcome that could be used consistently across studies, but it is limited because differences are unlikely to be observed for at least 10–15 years, and few of the ablative studies had such length of follow-up. Dysfunctional and quality of life outcomes can be used for comparison, but were limited for the reasons given previously. Need for further (systemic) treatment may also be used to contrast all therapies, but again this was rarely reported in the studies. All of these issues contributed to the review providing little information on the comparative effectiveness of AS and active treatment.

Cost-effectiveness