Collagenase clostridium histolyticum for the treatment of Dupuytren's contracture: systematic review and economic evaluation

Introduction

Dupuytren’s disease is a benign, slowly progressive, fibroproliferative condition that affects the palmar and digital fascia in the hand (the ‘bands’ that anchor the skin of the palm). 1 The disease is common, costly and can impact on quality of life. 2–4 It is characterised by thickening of the palmar skin and by the formation of nodules, firm painless masses fixed to the skin and the deeper fascia. Nodules usually precede the development of fibrotic cords. As the disease progresses, cords gradually contract, leading to progressive flexion deformities in the fingers, particularly of the metacarpophalangeal (MCP) and/or proximal interphalangeal (PIP) joints. 5,6 Joint contractures and deformities are usually painless7 but are associated with considerable functional impairment which, in turn, impacts on activities of daily living and ability to work. 4

There is currently no cure for Dupuytren’s disease and the goal of treatment is to restore hand function. 7 Management of the disease is dependent on disease progression and degree of deformity, and not all people with Dupuytren’s disease seek or require treatment. 1,8 Surgery remains the treatment of choice for severe contractures and some cases of moderate symptoms. 1,8,9 Surgery, however, cannot be considered a definitive cure for Dupuytren’s disease, as contracture may recur in operated digits or in previously uninvolved areas of the hand. Moreover, perioperative and/or postoperative complications are relatively common after surgical treatments. 2,3,10 Recently, collagenase clostridium histolyticum (CCH) (Xiapex^®, Pfizer Ltd) has been proposed as a non-operative, clinically viable alternative to surgery in some patients and has been licensed for the treatment of Dupuytren’s contracture (DC) in adults. 11,12

The aim of this appraisal is to assess the current evidence on the clinical effectiveness and cost-effectiveness of CCH as an alternative to surgery for treatment of adults presenting with DC with a palpable cord.

Aetiology, pathophysiology and clinical presentation

Although the aetiology of Dupuytren’s disease remains unknown,2 a clear genetic component13,14 involving autosomal dominant transmission with variable penetrance, has been established. 1,9,15

A number of risk factors have been reported to be associated with Dupuytren’s disease, such as alcohol consumption, smoking, diabetes, epilepsy, thyroid disorders and trauma, but none has been found to be definitively responsible for the condition. 16,17 A recent large cohort study conducted in France reported a significant dose–effect relationship between Dupuytren’s disease and age, diabetes, high alcohol consumption and long-term exposure to work-related vibration tools. 18 Smoking has also been associated with an increased risk of the disease, and the combination of smoking and high alcohol intake has been reported to increase the risk further. 19

The earliest signs of Dupuytren’s disease are changes in the skin, including loss of normal architecture and development of skin pits, caused by small vertical fibres (known as Grapow fibres) that connect the dermis to the palmar fascia. 1 The intermediate phase of the disease is characterised by an abnormal deposition of collagen. Initially, the collagen forms nodules (i.e. firm, painless masses fixed to the skin and deeper fascia) but it may thicken and lead to the formation of cords, which also adhere to the skin of the palm. 9,20 A cord may develop without prior formation of a nodule,9 although this remains contentious. 3 Further progression is marked by gradual contraction of cords, causing finger flexion deformities. 1,9 The MCP and PIP joints of the ring finger and the little finger are most commonly affected, but contractures may also occur in other fingers and the thumb. Nodules and cords are the defining features of DC (Figure 1). 9,21

FIGURE 1.

Hand affected by DC. Reproduced with permission from Donald Sammut, 2013 (www.donaldsammut.com).

When the cords contract, the condition is usually referred to as DC. Contractures of the MCP joints are caused by the pretendinous cord, which develops from the pretendinous band. Proximal interphalangeal joint contractures arise from the central, spiral or lateral cords. The lateral cord can also result in contracture of the distal interphalangeal joint. 9

The terms ‘Dupuytren’s disease’ and ‘Dupuytren’s contracture’ appear to be used interchangeably in the current literature even though some authors have stressed the importance of correct definitions. 16 In general, epidemiology studies tend to concentrate on Dupuytren’s disease, whereas experimental studies focus on DC.

Epidemiology and prognosis

Dupuytren’s disease occurs in all races, but is known to be more prevalent in Northern European Caucasians. The epidemiology of the disease has been extensively described and studies have been conducted in numerous countries around the world. Prevalence rates of Dupuytren’s disease vary widely according to age, population groups, geographical location and methods of data collection, with rates ranging from 0.2% to 56%. 22 In Europe, reports of prevalence range typically from 4% in the male population in England23 to 30% in the Norwegian population aged over 60 years,24 with the highest prevalence rate (56%) reported by a study conducted on a group of chronic epileptic patients in the UK. 25 DC has also been reported to be particularly prevalent in the north-east of Scotland. 25 Overall, approximately 2 million people in the UK are believed to suffer from Dupuytren’s disease. 1

The disease is more common in men than in women, with a ratio of approximately 6 : 1. 22 Men tend to present for treatment at a younger age (in their fifth decade) than women (in their sixth decade). Prevalence and incidence of the disease tend to increase with age and both sexes are affected in equal proportions after the age of 80 years. 26,27

The ring finger is most commonly affected. 28 Rates of bilateral disease vary between 17%29 and 59%28 with symmetrical involvement of fingers in many cases. 28 In unilateral disease, involvement of the right hand has been reported to be twice as likely as involvement of the left hand in a Norwegian population. 28 However, more recent studies in the UK and Germany have reported almost equal involvement of both hands. 30

Impact of health problem

Dupuytren’s disease impacts on patients’ quality of life and results in both psychosocial and physical consequences. 22 Increased finger flexion deformities have been associated with decreased hand function. 31 Impaired hand function can limit normal activities at home (e.g. washing and dressing), in the workplace (e.g. manual labour) and in recreational and social interactions (e.g. sports, shaking hands). 32,33

Although surgery is considered the mainstay of treatment for DCs, it does not guarantee success, and complications and recurrences are relatively common after surgery. 10 Surgery-related complications tend to increase according to the severity of the initial contracture and may be related to technical aspects of the surgical procedures (e.g. neurovascular injury, haematoma, infection) and to physiological characteristics of patients (e.g. stiffness, reflex symptomatic dystrophy). 9,29 A 20-year literature review reported complication rates for fasciectomy ranging from 4% to 39%, with a major complication rate of 16%. Pain and wound healing complications were more common after fasciectomy in people treated for primary disease, whereas in those with recurrent disease, necrosis and sensory abnormalities were more often reported. In studies directly comparing primary and recurrent disease, both digital nerve injuries and digital artery injuries were more common in patients with recurrent disease, but numbers were too small for drawing reliable conclusions. 10 A review of the literature that focused on the effects of fasciectomy and fasciotomy in European patients treated for Dupuytren’s disease showed that complications were experienced by approximately one-quarter of people undergoing fasciectomy and by one-fifth of those receiving fasciotomy. Overall, about 20% of patients who underwent fasciectomy or fasciotomy experienced a complication, including neurapraxia, nerve or arterial injury, infection, haematoma, pain and skin necrosis. 34

Recurrence of contracture following surgery is common and rates increase with severity of the initial contracture and with length of follow-up, regardless of the success of the surgery. 11,29 A systematic review of outcomes of surgery for primary disease reported recurrence rates of 0% to 71%, with follow-up ranging from 3 weeks to 13 years. A further review assessing the efficacy and safety of fasciectomy and fasciotomy in European patients with primary and recurrent disease showed an average recurrence rate of 39% after fasciectomy and 62% after fasciotomy at a median follow-up time of about 4 years. 34 In addition, a systematic review of surgical outcomes for DC found that recurrence rates for open partial fasciectomy (PF) ranged from 12% to 39% between 1.5 and 7 years and those for percutaneous needle fasciotomy (PNF) ranged from 50% to 58% between 3 and 5 years. 34 In the current literature, there is clear heterogeneity in recurrence rates reported after surgery.

Definitions of recurrence are not consistent across studies and some investigators do not define recurrence at all. It is, therefore, rather challenging to compare recurrence rates across studies with any degree of confidence. 35,36

Management of the disease

In the UK, the majority (76%) of DC diagnoses are made by the general practitioner (GP). Most patients (82%) are then referred by the GP, with around half (49%) of referrals being to a hand specialist. Post-operative management is carried out mainly by a physiotherapist (in 59% of cases) or the treating surgeon (in 25% of cases). 3

The British Society for Surgery of the Hand (BSSH) clinical classification of Dupuytren’s disease is displayed in Table 1.

Various algorithms for managing DC have been proposed in the literature. 1,2,8

Percutaneous needle fasciotomy is recommended, if appropriately trained, for the treatment of moderate MCP joint contracture. Although the marketing authorisation for CCH is ‘for the treatment of DC in adult patients with a palpable cord’, the BSSH tentatively recommends collagenase injections for adults with moderate contractures. 37 It is worth mentioning that this guidance was issued before collagenase was licensed for use in the UK. The BSSH has also recently withdrawn all previously published evidence-based guidelines with the intent to review them. No new updated information is currently available from the BSSH.

There is no established cure for Dupuytren’s disease and the therapeutic goal is to restore hand function. 6,38 Surgery has been the treatment of choice for severe contractures and for some people with moderate contractures. 1,2,9,32 There are no definitive indications for treatment but contraction of the MCP joint of > 30° is a commonly accepted indication. 9 Indications for surgery of PIP joints vary from contractures of 15°9 to any degree of contracture. 39 Either way, earlier intervention for PIP joints is common to guidance across the board. 3 In addition, the Hueston table top test40 is a simple assessment of the need for surgical intervention. Inability to place the hand flat on a table top indicates a positive test.

Various non-surgical interventions have been proposed for Dupuytren’s disease, including local vitamin E cream, local steroid injections, splinting and ultrasonic therapy, but these have been mostly abandoned because they are ineffective or inferior to surgery. 11,41 There is some evidence that radiation therapy may be effective in slowing down the progression of the disease without serious safety concerns. However, the use of radiation therapy in clinical practice in the UK is very limited and requires special arrangements for clinical governance, consent and audit or research. 42

Variation in services and/or uncertainty about best practice

There are currently no definitive guidelines regarding the type of surgery used to treat DC. Fasciectomy remains the preferred choice for the treatment of DC. However, none of the available surgical procedures has proved completely effective and complications and recurrences are common. The choice of surgery is influenced by a number of factors, including the age of patient, patient preference, comorbidities, severity of disease, as well as surgeon preference and expertise. 6,43 In 2008, a large survey was conducted among a sample of 687 orthopaedic/plastic surgeons across 12 European countries (including the UK) to estimate the geographical variation in the number of interventions for Dupuytren’s disease performed during the previous 12 months. Ninety five per cent of surgeons used fasciectomy, 70% used fasciotomy, 38% used dermofasciectomy and 35% used PNF. 43,44 In Europe, between 4% and 12% of patients with DC are treated with PNF, depending on the country.

Relevant national guidelines

Currently, there are two sets of National Institute for Health and Care Excellence (NICE) guidelines for the treatment of Dupuytren’s disease/contracture. The interventional procedure guidance 43 ‘Needle fasciotomy for Dupuytren’s contracture’ was published in February 2004. 45 The guideline states that needle fasciotomy, a procedure that can be repeated, offers a short-term reduction of contracture, with a recurrence rate of about 50% at 5 years. The guidance further indicates that people with less severe disease and/or MCP contracture are likely to benefit most from the procedure, although it is not clear whether such additional benefit is in terms of reduction of contracture or time to recurrence. Needle fasciotomy is reported to be less efficacious in the longer term than open surgery but is associated with lower morbidity and faster recovery times than open procedures. Complications of the procedure, including skin splitting, pain and nerve injury are acknowledged in the guideline, but with a rate of ≤ 1%. NICE conclude that needle fasciotomy is recommended especially for older people for whom major surgery is not an option.

The IPG368 NICE guideline was published in November 2010 and focuses on the use of radiation for early Dupuytren’s disease. 42 The guideline states that radiation offers correction of contraction, restoration of hand function and avoidance of future surgery but with a theoretical risk of radiation-induced cancer in the long term. The guideline concludes that the current efficacy and safety data for radiation therapy for early Dupuytren’s disease are limited and recommends the use of the procedure only with special arrangements for clinical governance, consent, and audit or research.

The BSSH guidelines classify Dupuytren’s disease as mild, moderate or severe. The guidelines recommend observation for mild disease, together with reassurance and a follow-up assessment between 6 months and 12 months. 37 Limited fasciectomy (LF) and PNF for MCP joints contracture (when the operating clinician has received appropriate training) are recommended for moderate disease, whereas fasciectomy (LF or dermofasciectomy) is recommended for severe disease. Collagenase injections are recommended only tentatively for moderate and severe disease owing to the lack of long-term safety data at the time of the recommendations.

Collagenase clostridium histolyticum

Collagenase clostridium histolyticum is a novel non-surgical treatment which has a UK marketing authorisation for the treatment of DC in adults with a palpable cord. It is marketed in Europe as Xiapex^® (Pfizer Ltd) and in the USA as Xiaflex^® (Auxilium Pharmaceuticals, Inc.). 12 The collagenase preparation is a lyophilised product for parenteral administration which comprises two discrete types of collagenase in a defined mass ratio. These collagenases are representative of the two major collagenase classes (class I and II) produced by Clostridium histolyticum. Both collagenases effectively cleave interstitial collagen but at different sites on the molecule. Class I collagenase cleaves the terminal ends of the collagen and class II collagenase cleaves internal sections of collagenase. 11,46,47 According to the summary of product characteristics, it is administered by intralesional injection, with the recommended dose being 0.58 mg. 12

The cord is adjacent to the flexor tendons and precise depth of injection is of paramount importance11 to avoid injection of the flexor. 48 The injection is administered in a slow and steady fashion to prevent the liquid being forced through the cord and into the deep fat surrounding the flexor tendons. 11 Approximately 24 hours after injection, the treated joint is manipulated to allow finger extension and encourage rupture of the cord. 7 This finger extension procedure may require the use of a local anaesthetic. If a satisfactory response has not been achieved, the injection and finger extension procedure may be repeated after approximately 4 weeks. Up to three injections per cord are recommended, at approximately 4-week intervals. Only one cord may be treated at a time. Where there are multiple contractures, each cord must be treated sequentially. The current clinical experience is limited to up to three injections per cord and up to a maximum of eight injections in total. 12 The collagenase procedure is, in some respects, similar to PNF; both procedures are minimally invasive, performed as outpatient procedures and involve the use of a needle.

Surgical interventions for Dupuytren’s disease

The most common type of surgery in the UK and throughout Europe is fasciectomy, which is the excision of the fascia from the affected area. 6 Fasciectomy can be limited to the excision of the diseased fascia, with the dissection carried no further than the PIP joint (limited/partial fasciectomy) or can involve the removal of all the palmar fascia (total/radical fasciectomy). Longitudinal or transverse skin incisions can be used. Longitudinal incisions are commonly closed with a Z-plasty, whereas zigzag incisions are closed directly (Bruner technique). The procedure requires the use of a general or regional anaesthetic and can be carried out as day surgery or inpatient admission. 1 The overlying skin is normally preserved during the procedure but skin fixation to the cord can occur. 49 Dermofasciectomy involves excision of both the fascia and the overlying skin, with lost tissue being replaced by a full-thickness skin graft. 49–51 Fasciotomy involves the simple division of the affected cord, by means of open surgery or percutaneously. 1 In open fasciotomy, a scalpel is used to cut the cord. 34,52 By contrast, PNF is a less invasive technique where the cord is sectioned with a 25-gauge needle mounted on a syringe. 21 PNF is recommended by NICE IPG43 and is considered particularly suitable for older people who are unfit for, or may not tolerate, major surgery. 45 Amputation may be an option in severe cases with later presentation or recurrence of the disease. Hand therapy and the use of splints are often required after surgery in order to maximise and maintain reduction of the contraction and improvement in finger position.

In Europe, factors that influenced the surgeons’ decision to perform fasciectomy or dermofasciectomy were consistent across countries: contracture of MCP or PIP joints > 45°, recurrent contracture and high expectations of success. For PNF, defining factors were more variable but MCP flexion of < 20° was a crucial factor. 43

Identification of important subgroups

No subgroups were specified in the final scope for this appraisal. However, consideration of people with moderate contracture and with severe contracture as relevant subgroups may be informative, as these groups are likely to be offered different surgical treatments. In addition, it might be advantageous to consider treatment of MCP joints versus PIP joints as subgroups, as each type of joint responds differentially to treatment. 29

Population

The population considered for this assessment is adults (18 years of age and older) with DC with a palpable cord. CCH is licensed ‘for the treatment of DC in adult patients with a palpable cord.’ Nevertheless, BSSH current guidelines recommend reassurance/observation for mild cases and collagenase injections for adults with moderate or severe contractures.

Population

The manufacturer’s submission focused on adults with moderate or severe DC with a palpable cord. This is not in line with the final scope issued by NICE, which does not specify any severity of disease and considers suitable for inclusion ‘adults with Dupuytren’s contracture with a palpable cord’. However, as people with mild contracture would not be eligible for treatment with collagenase in any case (according to the BSSH guidelines), the manufacturer’s approach appears to be reasonable. Inclusion criteria for both CORD I55 and CORD II56 trials specify MCP joint contractures of 20–100° or PIP joints of 20–80°. The manufacturer’s submission reports 14/167 (8.4%) participants with mild disease in the subgroup of participants with two or fewer affected joints. The total number of participants with mild disease is unclear owing to total contracture index being the main reported measure of contracture.

Intervention

The intervention specified in the manufacturer’s submission (i.e. CCH) matches that of the NICE final scope.

Comparators

The NICE final scope states that the comparators are surgical treatments, including fasciectomy, open fasciotomy, needle fasciotomy. The manufacturer’s submission differs from the NICE final scope in that only fasciectomy was considered a suitable comparator. The manufacturer justifies this decision by arguing that fasciectomy is the main treatment recommended by the BSSH and is the most commonly used procedure in England. Furthermore, they maintain that the remaining specified treatments are either not recommended by the BSSH (i.e. open fasciotomy) or used in only a small proportion of patients with moderate to severe disease (i.e. PNF). The exclusion of PNF is questionable. PNF is still performed by some surgeons in some clinical departments in the NHS and is suitable for less severe disease and for elderly people who are unsuitable for, or may not tolerate, surgery. 45 In addition, among all surgical procedures, PNF shows more similarities with collagenase (both are minimally invasive procedures that do not require in-hospital stay).

Outcomes

The outcomes considered by the manufacturer were those specified in the NICE final scope (i.e. reduction of contracture, time to return to normal function, recurrence of contracture, adverse effects of treatment and health-related quality of life).

Inclusion and exclusion criteria

Data extraction strategy

Two reviewers (MC and CR) were involved in the screening process. Each reviewer screened half of the titles and abstracts identified by the search strategies. To ensure consistency, the first 20 citations were double-screened by both reviewers. Any disagreements were resolved by consensus. All potentially relevant reports were retrieved in full and assessed independently by the same two reviewers. Any disagreements were resolved by consensus or referred to a third author. A sample of the full-text screening form is shown in Appendix 2.

A data extraction form was developed and piloted for the purpose of this appraisal. From each included study, one reviewer (MC, CR or MB) collected information on journal name, publication year and status, study design, recruitment method, setting, characteristics of participants, characteristics of interventions and outcome measures. Data extraction was double-checked by a second reviewer with the exception of a Portuguese-language paper,57 which was not double-checked. Any disagreements were resolved by discussion or arbitration by a third author.

Critical appraisal strategy

The standard Cochrane risk of bias tool was used to assess the risk of bias in randomised trials. 58 Included RCTs were assessed using Cochrane’s tool for assessing risk of bias. A sample form is presented in Appendix 3. Two reviewers (CR and MC) independently assessed risk of bias within each included RCT, based on the following domains: sequence generation, allocation concealment, blinding, incomplete outcome data and selective outcome reporting. Individual outcomes were categorised as being at high, low or unclear risk of bias. Any disagreements between reviewers were resolved by consensus.

An adapted version of the risk of bias tool recommended by the Cochrane Non-Randomised Studies Methods Group was used for non-randomised comparative studies (see Appendix 4). 58 The main confounders were identified a priori from data derived from the existing epidemiological literature and from consultation with the advisory panel convened for this assessment. A study was judged to be at high risk of bias if any of the identified confounders were imbalanced (e.g. type of joint, degree of baseline contracture).

Risk of bias of case series was assessed using a modified version of a 17-item checklist previously developed by our research team (see Appendix 5). The checklist was originally adapted from several sources and developed through a partnership with the Review Body for Interventional Procedures (ReBIP) for NICE. 59–62 The case series tool assessed the following domains: bias and generalisability, sample definition and selection, description of the intervention, outcome assessment, adequacy of follow-up and performance of statistical analyses. Individual items were rated as ‘yes’, ‘no’ or ‘unclear’. A rating of ‘yes’ indicated a low risk of bias.

Two reviewers (MC and CR) independently assessed the risk of bias of all included primary studies. Any disagreements or uncertainties were resolved by discussion between the two reviewers. A third reviewer (MB) acted as an arbitrator where consensus could not be reached.

Methods of data synthesis

Results of each included study were tabulated for all outcomes with means reported for continuous outcomes and proportions for dichotomous outcomes. Where the same outcome was assessed by more than one included study, a quantitative synthesis of results was carried out using Review Manager software (Version 5.2.; Copenhagen: The Nordic Cochrane Centre, Cochrane, 2012). Heterogeneity between studies was assessed by visual inspection of forest plots and from Mantel–Haenszel chi-squared and I² statistics. Meta-analyses were carried out to estimate risk ratios (RRs) pooled across studies, with corresponding 95% confidence intervals (CIs). Where there was a large amount of heterogeneity between studies, random-effects models were applied using the inverse-variance method, otherwise fixed-effects models using the Mantel–Haenszel method were applied. Where SDs were not reported for a continuous outcome (and not subsequently supplied following a request to the authors), the values were imputed using data from other studies included in the meta-analysis, with this approach being tested in sensitivity analyses using differing values for the imputed SDs.

Number and type of studies included

The literature searches identified 720 potentially relevant citations and 502 conference proceedings. Eighteen ongoing trials were identified; the latest updates showed that two trials were not yet recruiting, one had finished recruiting but no results are available, 11 were still recruiting and four were ongoing but not recruiting. We selected and retrieved 187 reports for full-text assessment. Of these, 153 were subsequently excluded (Figure 2). We included a total of five RCTs (493 participants), published in six papers, comparing collagenase with placebo,55,56,63–66 three RCTs (334 participants), published in four papers, comparing various surgical procedures,67–70 two non-randomised studies (105 participants) comparing collagenase with surgery,71,72 five non-randomised studies (3571 participants), published in six papers, comparing various surgical procedures,44,57,73–76 and 15 collagenase case series (3154 participants) published in 16 papers. 48,75,77–90

FIGURE 2.

Flow diagram outlining the selection process.

Appendix 6 provides bibliographic details of all included and excluded studies.

Number and type of studies excluded

Forty-four reports were excluded because they failed to meet one or more of the pre-specified inclusion criteria with regard to type of study design, characteristics of participants, characteristics of intervention and comparators and outcomes measures.

Characteristics of the included studies

A summary of the baseline characteristics of included studies is reported in Table 3. An overview of the main characteristics of the included studies is presented in Table 4. Appendices 7 and 8 present the characteristics of the included collagenase and surgery studies, respectively.

Risk of bias of the included studies

Randomised controlled trials

Collagenase versus placebo

The four collagenase RCTs published in full55,56,64,65 were assessed for risk of bias. Secondary reports63,66 were not assessed. CORD I and CORD II55,56 were judged as being at low risk of bias for both sequence generation and allocation concealment. The risk of bias was unclear in the remaining RCTs. All RCTs were judged as being at low risk of performance bias as all successfully blinded participants and, with the exception of the CORD II trial,56 the health-care providers. For all RCTs, the risk of detection bias was unclear as it was not possible to determine with certainty whether outcome assessors were blinded but all performed intention- to-treat analyses. All collagenase RCTs were sponsored by pharmaceutical companies involved in the development of collagenase (Advance Biofactures Corporation,64 Biospecifics Technologies Corporation65 and Auxilium Pharmaceuticals)55,56 and, therefore, were judged to be at high risk of ‘other bias.’

Surgery versus surgery

The three surgery RCTs,67–69 were assessed for risk of bias. The secondary report by van Rijssen et al. 70 was not assessed. Two studies67,69 were judged as being at low risk of bias for both sequence generation and allocation concealment. Only one of these studies67 blinded participants, but neither blinded health-care providers. In one trial,69 it was not possible to determine with certainty whether outcome assessors were blinded and, therefore, the risk of detection bias was judged as ‘unclear’. Two studies did not blind outcome assessors. 67,68 Two studies performed intention-to-treat analyses. 68,69 This was unclear in one trial. 67

Summaries of the risk-of-bias assessments for the RCTs comparing collagenase versus placebo and for those comparing different surgical interventions are presented in Figures 3 and 4, respectively. Summaries of the individual study level assessments are provided in Appendices 9 and 10 for collagenase and surgery studies, respectively.

FIGURE 3.

Summary risk-of-bias assessment for collagenase RCTs.

FIGURE 4.

Summary risk-of-bias assessment for surgery RCTs.

Non-randomised comparative studies

Collagenase versus surgery

We assessed two collagenase versus surgery non-randomised studies. 71,72 Both were judged as being at high risk of bias. Participants in both studies were allocated to treatments by patient or physician preference, and the majority of outcomes were at high risk of bias as a result of confounding and performance and detection biases, including the lack of blinding procedures. Both studies were either funded by Auxilium Pharmaceuticals71 or noted financial affiliations between study authors and the company. 72

Surgery versus surgery

We assessed five non-randomised studies on different surgical procedures. 57,73–76 All studies were judged as being at high or unclear risk of overall bias. Only two studies74,76 were judged to be at low risk of bias for sequence generation and allocation concealment. The majority of outcomes were also at either high or unclear risk for confounding and blinding. Studies were generally at lower risk for attrition bias and selective reporting. One study73 was funded by Pfizer. The remaining studies did not report information on sponsors.

Summaries of the risk-of-bias assessments are presented in Figures 5 and 6. The results of individual study level assessments for non-randomised comparative studies are provided in Appendices 11 and 12.

FIGURE 5.

Summary risk-of-bias assessment for non-randomised comparative studies on collagenase vs. surgery. HRQoL, health-related quality of life.

FIGURE 6.

Summary risk-of-bias assessment for non-randomised comparative studies on different surgical procedures. HRQoL, health-related quality of life.

Case series studies

We assessed 11 collagenase case series. 48,78,79,84–91 Secondary reports81–83 and abstracts77,80 were not assessed. The majority of studies included a representative population, collected data prospectively, clearly defined the intervention, administered the collagenase injection in an appropriate setting by someone experienced at performing the procedure, and identified important prognostic factors. Three studies,78,90,91 however, included a mixed patient population (i.e. participants with primary disease and with disease recurrence). In five studies,48,85–87,89 it was unclear whether participants were at a similar point in their disease progression. In about half of the studies (54.6%), follow-up periods were not adequate (< 1 year) and the majority (72.7%) failed to provide information on the characteristics of participants who withdrew or did not complete follow-up. All studies were sponsored by Auxilium Pharmaceuticals48,78,79,84,86–91 or had a potential conflict of interest with the company. 85 A summary of the risk-of-bias assessment of the case series studies is presented in Figure 7 and the results of individual study level assessments are provided in Appendix 13.

FIGURE 7.

Summary risk-of-bias assessment for collagenase case series studies. N/A, not applicable.

Clinical success: all first joints

A total of 171/271 (63%) of all first joints treated with collagenase and 8/136 (6%) first joints treated with placebo achieved clinical success, defined as a reduction in contracture to 0° to 5° of normal, 30 days after the last injection (Figure 8). The difference between the groups was statistically significant (RR 10.21, 95% CI 5.29 to 19.69; p-value < 0.00001).

FIGURE 8.

Forest plot of proportion of all first joints achieving clinical success. df, degrees of freedom; M–H, Mantel–Haenszel.

Clinical success: first metacarpophalangeal joints

A total of 127/167 (76%) first MCP joints treated with collagenase achieved clinical success, compared with 6/87 (7%) joints treated with placebo (Figure 9). The difference between the groups was statistically significant (RR 10.27, 95% CI 4.88 to 21.65; p-value < 0.00001).

FIGURE 9.

Forest plot of proportion of first MCP joints achieving clinical success. df, degrees of freedom; M–H, Mantel–Haenszel.

Clinical success: first proximal interphalangeal joints

A total of 44 of 104 (42%) first PIP joints treated with collagenase and 2 of 49 (4%) joints treated with placebo achieved clinical success (Figure 10).The difference between the groups was statistically significant (RR 7.44, 95% CI 2.44 to 22.62; p-value = 0.0004).

FIGURE 10.

Forest plot of proportion of first PIP joints achieving clinical success. df, degrees of freedom; M–H, Mantel–Haenszel.

Clinical improvement: all first joints

Significantly more first joints treated with collagenase (207/248; 83%) than with placebo (15/124; 12%) achieved clinical improvement (Figure 11), defined as a reduction in contracture of 50% or more from baseline 30 days after the last injection (RR 6.90, 95% CI 4.28 to 11.12; p-value < 0.00001).

FIGURE 11.

Forest plot of proportion of all first joints achieving clinical improvement. df, degrees of freedom; M–H, Mantel–Haenszel.

Clinical improvement: first metacarpophalangeal joints

A total of 144 of 153 (94%) first MCP joints treated with collagenase achieved clinical improvement, compared with 10 of 80 (13%) first joints treated with placebo (Figure 12). The difference between the groups was statistically significant (RR 7.54, 95% CI 4.21 to 13.49; p-value < 0.00001).

FIGURE 12.

Forest plot of proportion of first MCP joints achieving clinical improvement. df, degrees of freedom; M–H, Mantel–Haenszel.

Clinical improvement: first proximal interphalangeal joints

A total of 63 of 95 (66%) first PIP joints treated with collagenase achieved clinical improvement compared with 5 of 44 (11%) first PIP joints treated with placebo (Figure 13). The difference between the groups was statistically significant (RR 5.85, 95% CI = 2.53 to 13.53; p-value < 0.0001).

FIGURE 13.

Forest plot of proportion of first PIP joints achieving clinical improvement. df, degrees of freedom; M–H, Mantel–Haenszel.

Mean change in range of motion from baseline for all first joints

Standard deviation for ROM at 30 days after last injection was not reported in the CORD I trial and not available from the author;55 we used the corresponding SD from CORD II. 56 This approach is in line with the Cochrane Handbook recommendations. 58 In addition, we conducted sensitivity analyses in which we varied the SD for both groups with no change in the results of the analysis (see Appendix 14 for full details of sensitivity analyses). 55

Figure 14 shows that the mean change in ROM from baseline was significantly greater for first joints treated with collagenase than for those treated with placebo (mean difference 31.84, 95% CI = 28.39 to 35.29; p-value < 0.00001).

FIGURE 14.

Forest plot of mean change in ROM from baseline for all first joints (using CORD II SD for both studies). df, degrees of freedom; IV, inverse variance; M–H, Mantel–Haenszel.

Mean change in first joint contracture from baseline

The CORD I trial did not report the SD for the mean change in contracture from baseline. Therefore, we used the corresponding value from the CORD II trial and performed sensitivity analyses, which showed no difference in results (see Appendix 14 for full details of sensitivity analyses). 56

Figure 15 shows that first joints treated with collagenase achieved a significantly greater change in contracture from baseline than joints treated with placebo (mean difference 65.27, 95% CI 52.06 to 78.48; p-value < 0.00001).

FIGURE 15.

Forest plot of mean change in first joint contracture from baseline (using CORD II SD for both studies). df, degrees of freedom; IV, inverse variance; M–H, Mantel–Haenszel.

Adverse events: participants experiencing at least one adverse event

A total of 265 of 272 (97%) joints treated with collagenase experienced at least one adverse event compared with 39 of 137 (28%) joints treated with placebo (Figure 16). The difference between groups was significant (RR 2.49, 95% CI 1.13 to 5.50; p-value = 0.02).

FIGURE 16.

Forest plot of proportion of participants experiencing at least one adverse event. df, degrees of freedom; M–H, Mantel–Haenszel.

Adverse events: peripheral oedema

A total of 183 of 249 (73%) participants treated with collagenase and 6 of 125 (5%) of those treated with placebo experienced peripheral oedema (Figure 17). This difference between the groups was significant (RR 15.23, 95% CI 6.97 to 33.29; p-value < 0.00001).

FIGURE 17.

Forest plot of proportion of participants experiencing peripheral oedema. df, degrees of freedom; M–H, Mantel–Haenszel.

Adverse events: contusion

Figure 18 shows that significantly more participants treated with collagenase (137/249; 55%) experienced contusion than those (4/125; 3%) treated with the placebo (RR 14.09, 95% CI 4.20 to 47.30); p-value < 0.0001).

FIGURE 18.

Forest plot of proportion of participants experiencing contusion. df, degrees of freedom; IV, inverse variance; M–H, Mantel–Haenszel.

Adverse events: pain in extremity

A total of 88 of 249 (35%) participants treated with collagenase and 7 of 125 (6%) of those treated with placebo experienced pain in extremity (Figure 19). The difference between groups was statistically significant (RR 6.26, 95% CI 3.00 to 13.09; p-value < 0.00001).

FIGURE 19.

Forest plot of proportion of participants experiencing pain in extremity. df, degrees of freedom; M–H, Mantel–Haenszel.

Adverse events: injection site pain

A total of 106 of 272 (39%) participants treated with collagenase and 13 of 137 (9%) of those who received placebo suffered from injection site pain (Figure 20). The difference between groups was statistically significant (RR 3.49, 95% CI 1.48 to 8.27; p-value = 0.004).

FIGURE 20.

Forest plot of proportion of participants experiencing injection site pain. df, degrees of freedom; IV, inverse variance; M–H, Mantel–Haenszel.

Adverse events: injection site haemorrhage

The risk of experiencing injection site haemorrhage was significantly higher for patients receiving collagenase than for patients receiving placebo (Figure 21). In total, 95/249 (38%) participants treated with collagenase and 4 of 125 (3%) of those who received placebo had an injection site haemorrhage (RR 10.70, 95% CI 4.26 to 26.91; p-value < 0.00001).

FIGURE 21.

Forest plot of proportion of participants experiencing injection site haemorrhage. df, degrees of freedom; M–H, Mantel–Haenszel.

Adverse events: injection site swelling

Figure 22 shows that the incidence of injection site swelling was significantly higher among participants treated with collagenase (59/249; 24%) than among those (7/125; 6%) treated with placebo (RR 4.18, 95% CI 1.98 to 8.81; p-value = 0.0002).

FIGURE 22.

Forest plot of proportion of participants experiencing injection site swelling. df, degrees of freedom; M–H, Mantel–Haenszel.

Adverse events: tenderness

Figure 23 demonstrates that the risk of experiencing tenderness was significantly greater among participants treated with collagenase (60/249; 24%) than among those (0/125; 0%) treated with placebo (RR 18.99, 95% CI 2.21 to 163.10; p-value = 0.007).

FIGURE 23.

Forest plot of proportion of participants experiencing tenderness. df, degrees of freedom; IV, inverse variance; M–H, Mantel–Haenszel.

It is worth noting that none of the participants treated with placebo subsequently experienced tenderness compared with almost one-quarter of those treated with collagenase.

Adverse events: pruritus

A total of 28 of 272 (10%) participants treated with collagenase and 1 of 137 (< 1%) treated with placebo experienced pruritus (Figure 24). The difference between groups was statistically significant (RR 7.35, 95% CI 1.77 to 30.43; p-value = 0.006).

FIGURE 24.

Forest plot of proportion of participants experiencing pruritus. df, degrees of freedom; M–H, Mantel–Haenszel.

Adverse events: lymphadenopathy

The risk of lymphadenopathy was significantly higher after collagenase than after placebo (Figure 25). A total of 40 of 272 (15%) participants treated with collagenase and 0 of 137 (0%) of those treated with placebo experienced lymphadenopathy (RR 14.10, 95% CI 2.82 to 70.40; p-value = 0.001).

FIGURE 25.

Forest plot of proportion of participants experiencing lymphadenopathy. df, degrees of freedom; M–H, Mantel–Haenszel.

Adverse events: axillary pain

Few participants treated with collagenase (15/249; 6%) and none of those who received placebo (0/125; 0%) reported axillary pain (Figure 26). The difference between groups was statistically significant (RR 7.98, 95% CI 1.08 to 59.00; p-value = 0.04).

FIGURE 26.

Forest plot of proportion of participants experiencing axillary pain. df, degrees of freedom; M–H, Mantel–Haenszel.

Adverse events: injection site vesicles

The incidence of injection site vesicles was low and similar in both groups (Figure 27). Six of 249 (2%) participants treated with collagenase and 1/125 (< 1%) treated with placebo developed injection site vesicles (RR 2.16, 95% CI 0.37 to 12.55; p-value = 0.39).

FIGURE 27.

Forest plot of proportion of participants experiencing injection site vesicles. df, degrees of freedom; M–H, Mantel–Haenszel.

Serious adverse events

Four serious adverse events were reported among participants treated with collagenase: one case of complex regional pain syndrome and two cases of tendon rupture in CORD I,55 and one case of flexion pulley rupture in CORD II. 56

Appendix 15 presents a summary of outcomes, including adverse events, for all included collagenase studies.

Summary

Evidence for the assessment of the effects and safety of collagenase injections for the treatment of DC was derived primarily from two RCTs, CORD I and CORD II, with a total of 374 participants. 55,56 Data from a third RCT with a total of 35 participants65 were included in the analyses, where possible. Table 5 presents an overview of the meta-analyses findings.

Joints treated with collagenase were significantly more likely to achieve clinical success (i.e. a reduction of contracture to < 5° within 30 days after the last injection) and clinical improvement (i.e. reduction in contracture of 50% or more from baseline) than those treated with placebo, with MCP joints more likely to improve than PIP joints. Similarly, significantly more joints treated with collagenase showed a change in contracture and an increase in ROM from baseline than those treated with placebo. Nevertheless, joints treated with collagenase were significantly more likely to experience at least one treatment-related adverse event and a significantly higher number of adverse events than those treated with placebo, with the exception of injection site vesicles, which had an overall low incidence. Adverse events were generally mild or moderate, with the exception of four serious adverse events (one case of complex regional pain syndrome, two cases of tendon rupture and one case of flexion pulley rupture). The two largest trials (CORD I and CORD II)55,56 reported that none of the participants experienced a recurrence during the 90-day study period.

Summary of adverse events in collagenase case series studies

Two flexor tendon ruptures were reported in the 15 collagenase case series studies: Coleman et al. 79 reported a flexor tendon rupture 4 months after treatment and McMahon et al. 85 reported a flexor tendon rupture in a ‘small finger’ PIP joint. In addition, Coleman et al. reported a pulley rupture and a participant with hand pain requiring hospitalisation. 79 No other studies reported any treatment-related serious adverse events. Peimer et al. reported no new long-term or serious adverse events in the 3-year follow-up of the CORDLESS study. 87 Overall, 30% of CORDLESS participants experienced 370 mild or moderate adverse events.. Two studies did not report adverse events. 77,89 Another study reported that minor events but no major adverse events occurred, even though full details were not provided. 48 Five studies reported that all participants78,79 or the majority of participants (98%,81 87%,91 97%90 experienced at least one adverse event. Contusion was a frequently experienced adverse event, ranging from 60%90 to 100% of participants. 78 Reported rates of lymphadenopathy were generally low, ranging from < 1%86 to 7%90 of participants. Frequencies of other adverse events were more variable. Pruritus, for example, ranged from < 1%86 to 42% of participants. 78 Reported rates of peripheral oedema ranged between < 1%86 and 100% of participants. 78 Skin-related adverse events were also common. Skin tears ranged from < 1%86 to 85% of participants;80 injection site pain from 25%78 to 42%. 90

In summary, two collagenase case series reported a total of four serious adverse events: two flexor tendon ruptures, one pulley rupture and one episode of pain requiring hospitalisation. Mild or moderate adverse events were common, with some studies reporting that all participants experienced at least one adverse event. Common adverse events were contusion, peripheral oedema and skin-related events.

CORDLESS study: findings at the 5-year follow-up

The manufacturer’s submission provides 5-year data for the CORDLESS study, an open-label study evaluating participants who were enrolled in previous collagenase clinical trials. Forty-seven per cent of the previously successfully treated joints (291/623) recurred and 16.9% (105/623) had further medical or surgical interventions by 5 years.

Cost-effectiveness evidence

The manufacturer conducted two literature reviews. The first was a pragmatic literature review to identify previous economic evaluations of CCH. They identified only two cost-effectiveness papers from the pragmatic literature search. The searches for economic evaluations were undertaken in PubMed and in the National Health Service Economic Evaluation Database (NHS EED). No further details, in terms of search methods, key words or dates, were provided. The second was a systematic literature review conducted initially in 2011 and subsequently updated in August 2013 (see section 5.2 of the submission). This review had two primary objectives: (1) to assess the short- and long-term efficacy and safety of treatments for DC; and (2) to determine the possibility of performing an indirect comparison between injectable CCH and surgical interventions. The review did not identify any studies drawing a comparison between injectable CCH and surgical interventions. It was, therefore, impossible for the manufacturer to compare directly current therapies. Thus, they opted to conduct a cost-minimisation analysis rather than a cost–utility analysis.

The two papers identified in the pragmatic literature search were authored by Chen et al. 100 and Baltzer and Binhammer101 and were published in 2011 and 2013, respectively. We also identified these studies as part of our systematic literature review of economic evidence. Further details of these studies are provided (see Systematic review of existing cost-effectiveness evidence). In brief, Chen et al. developed an expected-value decision-analysis model, with a 20-year time horizon, to examine the cost-effectiveness of open PF, PNF and CCH injection compared with no treatment in a US setting. 100 A societal perspective was adopted. Quality-of-life (utility) weights were estimated from a utility survey constructed by the authors and specific to people with DC. Baltzer and Binhammer assessed the cost-effectiveness of percutaneous needle aponeurotomy (PNA) versus PF versus injectable CCH in a Canadian setting. 101 A 15-year time horizon, expected-value decision-analysis model was developed for patients undergoing primary treatment for DC affecting a single finger. The analysis was conducted from a societal perspective and utility weights were derived from previously published literature. 100 Both Chen et al. and Baltzer and Binhammer identified PNF as the most cost-effective treatment strategy. Overall, similar conclusions were reported in both studies, in that a non-surgical approach to DC is preferred and the cost of CCH needs to reduce significantly if it is to offer a cost-effective alternative to surgery.

As highlighted by the manufacturer in their submission, these studies display a number of limitations. First, data were derived from different studies; however, no adjustment was conducted to account for the heterogeneity across studies, in particular for the differences related to patient characteristics. Second, a standard gamble approach to derive utility weights may not fully capture the utility implications, especially if respondents were not willing to trade an additional risk of death. Furthermore, these values were applied for the rest of the time horizon within the models. However, as DC is not life-threatening, it may be challenging to capture accurately differences in quality of life associated with different health states using this method of estimation (i.e. respondents may have found the choice task difficult as it requires them to weigh up life lived in a chronic health state with certainty against a risky alternative of either death or return to full health).

Critique of the manufacturer’s economic evaluation

The manufacturers submitted a de novo cost-minimisation analysis comparing CCH and fasciectomy. The adopted perspective of the analysis was direct NHS and Personal Social Services (PSS) costs. A 5-year time horizon was adopted, with costs discounted at 3.5% per annum. This analysis was conducted for patients suffering from moderate or severe disease with two or fewer joints affected in the same hand. The rationale for choosing this subset of patients was that CCH is more likely to be administered to patients with fewer joints affected. Nevertheless, for all patients within the CORD I and CORD II trials, the mean number of joints affected per patient treated with CCH was 3 and 3.4, respectively. Furthermore, the manufacturer’s analyses were restricted to fasciectomy, as PNF was not judged to be a suitable comparator. Table 9 shows the findings of the comparison between the manufacturer’s economic submission and the NICE reference case. 102

Model structure

The manufacturer’s model structure adopts an annual cycle length. There are two phases in the model structure: initial treatment and recurrent disease. The patient enters the model and is treated with either CCH or fasciectomy (Figure 28). If two joints are affected in the CCH arm, a first joint is selected and treatment begins on this joint. Fasciectomy occurs in the initial treatment only, as a number of joints can be treated at one time. In the CCH arm, patients could receive a maximum of three cycles of treatment. If, after either strategy, the treatment was unsuccessful, it was considered a treatment failure and the patient accumulated no further costs. This is a potentially problematic assumption if the success rates between treatment options vary, and if a patient would go on to have further treatment if initial treatment failed. There is evidence that patients do indeed receive second-line treatment (Mr David Lawrie, Woodend Hospital, 2014, personal communication), and the model we have developed for this assessment takes this into account. If the strategies were successful, patients entered a treatment success health state where they incurred a probability of recurrence. Success rates were assumed equivalent for both strategies. This is a further problematic assumption; there is no evidence from non-inferiority RCTs to support the assumption of equivalence and, therefore, the method of cost-minimisation analysis is not supported by existing evidence. Moreover, the model could potentially bias the results favouring the strategy with a higher proportion of treatment failures. If the patient experienced recurrence, they were eligible for further treatment either in the form of fasciectomy or CCH, with the manufacturer assuming an equal distribution of treatments following recurrence between the two arms.

FIGURE 28.

The model structure. Source: manufacturer’s submission.

Population

Patient characteristics are drawn from CORD I and CORD II, two prospective, randomised, placebo-controlled trials with 90-day double-blind follow-up. CORD I enrolled 308 patients with a mean total contracture index of 149.1 in the CCH arm and 149.3 in the placebo arm. The CORD II study enrolled 66 patients with a mean total contracture index of 174.7 in the CCH arm and 150.1 in the placebo arm. The primary outcome for both studies was reduction of contracture to 0–5°, 30 days after the last injection was administered. The manufacturer’s model included patients with moderate (37.91%) and severe (62.09%) disease, with a maximum of two joints affected in the same hand. This is a subgroup of patients included in the CORD I and CORD II studies.

Intervention and comparators

The manufacturer’s submission focused exclusively on fasciectomy as the relevant comparator, deviating from the NICE final scope. The BSSH recommend PNF as an alternative to surgery for patients with moderate disease. 37 Advice from our clinical expert indicates that needle fasciotomy is an appropriate starting point for the treatment of Dupuytren’s disease and is currently used in clinical practice in the UK (Mr David Lawrie, personal communication). The economic analyses identified by the manufacturer in their literature review, have, despite their methodological limitations, identified PNF as the most cost-effective treatment strategy. Considering that the manufacturer’s analysis focused on patients with both moderate and severe diseases, PNF might have been an appropriate comparator.

Resource identification, measurement and valuation

The costs included in the manufacturer’s submission are provided in Tables 10 and 11. The model includes 1.6 injections of CCH per affected joint. This is the mean number of injections administered per joint from the post hoc analysis of the CORD I and CORD II studies and assumes that 1.4 joints are treated. Physiotherapy costs have not been included in the cost calculation of CCH. One session of physiotherapy is usually required following treatment with CCH. The manufacturer has also costed a total of 9.4 hand therapy appointments after fasciectomy. This appears to be an overestimation. Advice from our clinical expert indicates that the average number of follow-up appointments after fasciectomy is five. Table 10 provides a summary of the CCH costs used in the manufacturer’s submission. Table 11 shows the comparable costs used for the assessment of fasciectomy.

Perspective, time horizon and discounting

For costs, the perspective adopted was that of the NHS and PSS. As the chosen analysis was a cost-minimisation, health benefits were not included. The time horizon was 5 years, which mirrors the length of follow-up of the CORDLESS study. Costs were discounted at an annual rate of 3.5%.

Treatment effectiveness

The overall success rate (63.9%) was assumed equivalent between the CCH and fasciectomy arms. Patients who experienced a successful treatment entered a successfully treated health state, where they incurred a probability of recurrence. Patients who experienced a treatment failure dropped out of the model and incurred no further costs. Essentially, this means that patients successfully treated could, in theory, accumulate higher costs than those for whom treatment failed, and it is questionable whether this may reflect current clinical practice. The majority of recurrences were treated with fasciectomy (61.9%) and the rest (38.1%) with CCH. The distribution of treatments following recurrence was the same for both arms (CCH and fasciectomy).

The total recurrence costs over a 5-year time frame equated to £102. To achieve this figure, the manufacturer multiplied the success rate by the recurrence rate by the proportion treated for recurrent disease. This figure was subsequently multiplied by the unit cost of the procedure. There is a potential risk of double counting in this calculation (i.e. multiplication of the recurrence rate by the percentage of patients treated for a recurrence), which may underestimate the number of patients who are retreated.

Results

The base-case results (Table 12) identify CCH as cost saving by an amount of £943, compared with fasciectomy. The base-case analysis was based on an average of 1.4 joints treated. An identical cost of recurrence was applied for both treatments and thus the main driver in the observed difference in cost was the cost of surgery.

Sensitivity analysis

The manufacturer undertook a univariate sensitivity analysis to illustrate the impact on results of altering key model parameters. The number of CCH injections administered was the largest overall driver in the cost difference. However, when the maximum number of injections was administered, CCH remained cost-saving. The second largest driver of the cost difference was the proportion of patients receiving fasciectomy as inpatients; in the base-case analysis, 37.8% of patients were assumed to be inpatients. This assumption was taken from Gerber et al. ,6 who found that inpatient admission rates decreased from 58% in 2003–4 to 38% in 2007–8. The manufacturer demonstrated that the more patients treated in an outpatient setting, the less the cost saving experienced when comparing fasciectomy with CCH. Nonetheless, in the scenario in which 0% of patients received fasciectomy in an inpatient setting, CCH treatment still remained cost saving relative to fasciectomy. The manufacturer also included a scenario in which waste was included in the analysis; this occurs when a vial is opened but not entirely used and so must be discarded. Finally, a scenario was included in which the impact of 0% of patients receiving hand therapy was assessed. In all scenarios presented in sensitivity analyses, the manufacturer found CCH to be cost saving (Figure 29).

FIGURE 29.

Tornado diagram showing cost impact of varying model assumptions. Source: manufacturer’s submission. Ba, base; Hi, high; Lo, low.

Uncertainties related to the manufacturer’s submission

There are three main limitations in the case submitted by the manufacturer. First, there is an assumption that the success and recurrence rates between the treatment strategies are equivalent. However, equivalence, defined usually in terms of non-inferiority, has yet to be conclusively demonstrated (e.g. no RCTs comparing CCH vs. different forms of surgery have been conducted). Thus, at present, a cost-minimisation analysis may not be a valid approach to estimate cost-effectiveness, as it is entirely possible that the assumption of equivalence is not valid and that recurrence and success rates vary between CCH and fasciectomy.

Second, PNF is a viable alternative to both fasciectomy and CCH for patients with moderate disease. By excluding this relevant comparator from the cost-minimisation analysis, the manufacturer has deviated from the NICE final scope. This omission is far from negligible, as PNF is potentially a less expensive treatment option6 in terms of quantity and type of resources used at the point of delivery.

Third, there are concerns relating to some of the costing assumptions made by the manufacturer, namely (1) including 9.4 follow-up appointments for hand therapy following fasciectomy but 0 following the CCH treatment strategy; (2) assuming treatment failures would not incur any further costs; and (3) the possibility that recurrence costs may have been underestimated.

Systematic review of existing cost-effectiveness evidence

In addition to the literature review and cost-minimisation analysis prepared by the manufacturer, we conducted a comprehensive, structured review of the literature to identify relevant cost-effectiveness studies comparing the different interventions for DC.

A formal systematic review of existing economic evaluations relating to CCH injections or surgical treatments for adults with DC was performed. Highly sensitive search strategies were designed using appropriate combinations of both controlled vocabulary and text terms. The following databases were searched without language restrictions from 1995 onwards: NHS EED, HTA Database, MEDLINE, MEDLINE In-Process & Other Non-Indexed Citations, EMBASE, Science Citation Index and the Health Management Information Consortium (HMIC) database. Two previous cost-effectiveness studies were retrieved and were appraised against the British Medical Journal checklist for referees of economic analyses. The main findings of these studies are summarised in a narrative way and tabulated for comparison.

Summary of findings from identified studies

This literature review demonstrates that the potential cost-effectiveness of CCH is largely dependent on the cost of a series of CCH treatment. Chen et al. 100 stated that the entire CCH injection series of three injections would need to reduce to US$945 to offer a favourable cost per QALY. Baltzer and Binhammer101 drew a similar conclusion, although the price of CCH would need to be further reduced to CA$875 for a series of treatment to be considered cost-effective. It should be noted that a substantial variation in the cost per QALY of CCH was reported by both studies. Baltzer and Binhammer101 reported including patient-incurred costs in their analysis, whereas Chen et al. did not. 100 Furthermore, it is not clear if Chen et al. included the cost of treatment after recurrence or complications in their analysis. 100 Both studies appear to have derived the QALY values by simple multiplication of the health-state values derived by the time perspective of the model, with no further adjustment for probability of further treatment following recurrences and failures and subsequent incorporation of quality-of-life impacts (and changes in cost).

This approach does not fully account for changes in health states, which would be expected to occur over the time frame adopted. Also, the standard gamble method may not be able to account fully for the effect of differences in health state on utility, if respondents were not prepared to accept an additional risk of death in order to receive a greater probability of living in better health states.

Information regarding the Chen et al. 100 model is not fully reported in the paper and so it is difficult to understand properly the model structure and pathways. Baltzer and Binhammer illustrated the algorithm of the decision-analysis and discussed their model structure in more detail. 101 A summary of probabilities for the failure of treatment strategies and the complication development were tabulated. The main contrast between these studies is the manner in which the cost-effectiveness analysis was conducted: Chen et al. compared all treatment strategies to a status quo (no treatment), whereas Baltzer and Binhammer ranked strategies in order of increasing costs, with the least costly strategy becoming the first comparator in their analysis. 100,101

Furthermore, both studies incorporated different model time-frames. However, as a common starting age of 63 years was assumed in both models, with an average life-expectancy of 78 years, differing model time horizons should not account for any differences experienced in results. It is also important to note that, although univariate sensitivity analyses were conducted for both studies, neither study conducted a probabilistic sensitivity analysis to test the robustness of results to simultaneous changes in model parameters. It is widely accepted that to address uncertainty accurately within a model, both deterministic and probabilistic sensitivity analyses should be conducted. 102

Overall, similar conclusions were reported as both studies found that, firstly, a non-surgical approach to Dupuytren’s is preferred and, secondly, the cost of CCH needs to significantly reduce if it is to offer a cost-effective alternative to treatment. However, given the specificity of these studies to a US and Canadian health system in terms of quality-of-life valuation and costing technique, it is not clear if these results can be generalised to a UK setting. Table 13 provides a tabulated summary of the retrieved studies from the review. Our team’s searches did not identify any further cost-effectiveness studies in addition to those described in the manufacturer’s submission.

Economic assessment

Given the limitations of existing evidence to suggest equal effectiveness and, following the NICE scope and protocol, we developed a de novo decision analytical model to assess the cost–utility of CCH compared with PNF or LF in adults with DC.

Methods

A cost–utility Markov model was developed using TreeAge Pro (TreeAge Software, Inc., Williamstown, MA, USA, 2014). The alternative pathways were embedded in the Markov model simulating the downstream impact of treatment on recurrence and health-related quality of life. A Markov model considers patients in a discrete health state and allows the consequences of treatment strategies in terms of recurrence rates, health-related quality of life and costs to be captured for a particular patient population, over the adopted time horizon.

Costs incorporated in the model included those associated with treatment, complications and further treatment following a possible recurrence. Health-state utilities associated with pre and post treatment were incorporated in the model to calculate QALYs. Strategies were ranked incrementally in terms of their costs and ICER was calculated, applying a ceiling ratio of £20,000 per QALY. 102 Uncertainty was assessed by conducting deterministic and probabilistic sensitivity analyses. For the latter, the strategy with the highest net monetary benefit (NMB) was identified using the formula:

where NMB is the NMB of the strategy, E is the mean effect (in terms of QALYs), r^c represents the decision-makers’ maximum WTP for a QALY and C is the mean cost of the strategy. A value of £20,000 was applied for r^c.

Relevant patient population

The modelled cohort consisted of patients presenting with moderate or severe DC. Focus was on moderate and severe disease, as no surgical trials were identified that considered patients with mild disease only. Furthermore, BSSH guidelines do not recommend treatment for mild disease. 37 Although future updated guidelines are not expected to deviate from current recommendations not to provide treatment for mild disease, there is uncertainty regarding the extent to which these guidelines are implemented in clinical practice across the UK. For example, at the scoping workshop, consultees stated that CCH may be used to treat mild contracture in some cases. Thus, the decision problem is examining the value of CCH among patients who are considered suitable candidates for surgery. For the purposes of the model, the definition of patients who are eligible for surgery is derived from that of van Rijssen et al. ,70 who define the surgical cohort as ‘patients with total passive extension deficit of at least 30 degrees in any ray, the existence of a clearly defined palmar cord, and willingness to be considered for surgery’. The model is built on three key studies of effectiveness, described earlier in the systematic review, which identified one suitable head-to-head trial of LF versus PNF,70 in addition to the two placebo controlled randomised trials of CCH (CORD I and CORD II). 56,63–65 The mean age of the patients within these studies was 63 years, with a sex distribution of 84% male and 16% female; this was, therefore, the starting age and sex distribution of the modelled cohort. Within these trial populations, the mean number of joints affected per patient is three;55,70 therefore, our modelled cohort was assumed to have three joints affected. The base-case analysis was conducted using recurrence rates obtained from the literature. We carried out several analyses varying these recurrence rates to test the impact on cost-effectiveness. Analyses were also conducted where we varied the number of CCH injections required to complete a treatment course. In the base-case analysis, the number of injections required was 1.6 per joint; the impact of varying this to a lower estimate of 1 was also assessed. Further sensitivity analyses regarding the number of joints affected (1 or 2) was also conducted.

Care pathways

The care pathways were determined based on expert opinion and availability of data. Figure 30 shows the graphical representation of the model structure. For the purposes of the economic model, the representative cohort is patients with moderate or severe DC who are considered suitable for surgical treatment with LF, PNF, or CCH.

FIGURE 30.

Schematic representation of the economic model structure.

Following treatment with one of these procedures, patients can either experience treatment-related adverse events or no treatment-related adverse events. It was assumed that all patients would recover from any treatment-related complications experienced. Subsequently, patients entered either a treatment success or treatment failure health state. Treatment success and treatment failure rates were assumed to be equivalent for patients who did or did not experience complications. Patients who entered the treatment success health state incurred a risk of further disease recurrence. In the case of recurrence, patients could proceed with further treatment or, alternatively, could remain within the recurrence state, where they accumulated a lower health-related quality-of-life value. Patients who did not experience recurrence remained in the treatment success health state, in which they incurred no costs and accumulated a higher quality-of-life value. Finally, patients who do not achieve treatment success (i.e. have a treatment failure) could either transit to receive further treatment or remain in that health state, accumulating a lower health-related quality-of-life value.

Strategies to be evaluated

The strategies chosen for evaluation were selected based on the availability of data from the systematic review. It was not possible to obtain data for all strategies of clinical interest (e.g. for some patients with particular characteristics, dermofasciectomy may be considered a treatment option). Thus, the economic analysis focused on comparing CCH, PNF and LF. An important limitation of the available clinical effectiveness and cost-effectiveness literature is that no direct comparisons between CCH and any of the surgical treatments have been performed using a RCT study design, and no reliable indirect comparisons can be made owing to the lack of a common comparator. There are also differences in the patient characteristics between the surgical randomised studies and the CCH versus placebo trials and, additionally, probable unobservable differences in patient treatment history and study centres in terms of surgeon quality. The economic model is built from a naive indirect comparison and, as such, represents a departure from the NICE reference case. 102 The outputs from the model should, therefore, be considered cautiously.

Decision analytical model

The pathways were embedded in a Markov model developed to simulate the progression of treated DC. Thirteen events and states were used to model the care pathways: (1) initial treatment; (2) treatment success; (3) treatment failure; (4) recurrence; (5) second-line treatment; (6) treatment success following second-line treatment; (7) treatment failure following second-line treatment; (8) recurrence following second-line treatment; (9) third-line treatment; (10) treatment success following third-line treatment; (11) treatment failure following third-line treatment; (12) recurrence following third-line treatment; and (13) treatment complications. Patients who achieved a treatment success were modelled to progress towards a recurrent state based on observed recurrence rates from the literature. In the first cycle of the model, all patients were modelled to undergo treatment consisting of CCH, PNF or LF. Patients were then modelled to transit to a treatment success or treatment failure state for the subsequent model cycle. All patients faced a risk of recurrence, modelled based on rates observed for recurrence following the appropriate treatment strategy received. Each cycle length was 6 months and included a half cycle correction, such that probabilities of recurrence and costs of treatment and utility values were expressed within this cycle length. The age and sex-specific risk of death from all causes was also incorporated in the model based on interim UK life tables. 103

Retreatment for recurrence and treatment failure

The study by van Rijssen et al. reported the probability of further treatment following a recurrence. 70 Owing to a lack of available data, for the CCH and PNF strategies, we assumed that the probability of further treatment following a treatment failure after CCH and PNF was the same as that following a recurrence for CCH and PNF. Thus, patients who underwent further treatment following a recurrence after CCH or PNF were as likely to undergo further treatment following a treatment failure. We had no data for the probability of receiving PNF following CCH, and so we assumed that patients who opted for CCH would have been as likely to undergo PNF, had it been an option. Thus, patients who underwent further treatment had an equivalent probability of receiving either PNF or CCH. Further treatment options following PNF were based on clinical opinion, with patients modelled to receive LF (50%), PNF (40%) or CCH (10%). Patients experiencing treatment failure following LF (either on first-, second- or third-line treatment) did not receive any further treatment and thus entered a treatment failure health state (Mr David Lawrie, personal communication). Patients who experienced a recurrence following a successful treatment with LF could not receive any of the minimally invasive techniques and were therefore modelled to receive second- or third-line LF as appropriate (Mr David Lawrie, personal communication).

No evidence was identified on which procedure patients would choose following a second treatment failure or recurrence. For the base-case analysis, we therefore assumed that patients who proceeded for further (i.e. third-line) treatment following a second treatment failure (for CCH and PNF only) or recurrence would receive LF; this assumption was based on advice from our clinical expert. A sensitivity analysis was conducted in which all patients were modelled to undergo further treatment to test the effect on cost-effectiveness.

Table 17 reports the probabilities incorporated in the model of patients who received further treatment and those who did not. If patients failed after LF for any of the treatment strategies, they did not proceed for further treatment. Instead they entered a semi-absorbing state where they continued to accumulate a lower quality-of-life value for the remainder of the model. In the manufacturer’s model, 83% of patients did not undergo any further treatment. It is unclear whether this value is protocol driven and reflects current clinical practice; furthermore, it is much higher than the value reported by van Rijssen et al. 70 Therefore, for the base-case analysis, we assumed that the same proportion of patients who opted for further treatment in the LF strategy would also choose further treatment in the CCH arm. This assumption was made, as it was deemed to provide a fair balance between the manufacturer’s model where 83% did not undergo further treatment and the reported percentage of patients in the PNF strategy who did not undergo further treatment (27%) in the van Rijssen et al. study. 70 A sensitivity analysis was conducted to test the impact of this assumption, where the same proportion of patients in the CCH arm proceeded for further treatment as those in the PNF arm. Figure 31 outlines the treatment pathway used in the model for patients initially receiving either CCH or PNF. Figure 32 provides the pathway for patients initially receiving LF.

FIGURE 31.

Treatment pathway of patients with initial treatment of CCH or PNF.

FIGURE 32.

Pathway of patients with initial treatment of LF.

TABLE 17 - Patients undergoing further treatment and type of further treatment for recurrence and treatment failure

If patients failed after LF at any point in the model, they did not proceed for further treatment. Instead, they entered a semi-absorbing state where they accumulated no costs and a lower health-related quality-of-life value.

Procedure	Probability of further treatment	Distribution for PSA	Source
		Beta
CCH
No further treatment	0.60	Alpha: 6	Assumption: equivalent value as van Rijssen LF70
		Beta: 4
Further treatment	0.40		Assumption: equivalent value as van Rijssen LF70
Second-line treatment for CCH
CCH	0.19		Manufacturer’s submission and assumption
PNF	0.19		Manufacturer’s submission and assumption
LFa	0.62		Manufacturer’s submission
Third-line treatment for CCH
CCH	0		Manufacturer’s submission and assumption
PNF	0		Manufacturer’s submission and assumption
LFa	1		Manufacturer’s submission
PNF
No further treatment	0.27	Alpha: 12	van Rijssen et al., 201270
		Beta: 33
Further treatment	0.73		van Rijssen et al., 201270
Second-line treatment for PNF
CCH	0.10		Clinical opinion
PNF	0.40		Clinical opinion
LFa	0.50		Clinical opinion
Third-line treatment for PNF
CCH	0		Clinical opinion
PNF	0		Clinical opinion
LFa	1		Clinical opinion
LFa
No further treatment	0.60	Alpha: 6	van Rijssen et al., 201270
		Beta: 4
Further treatment	0.40		van Rijssen et al., 201270
Second- and third-line recurrence treatment for CCH
CCH	0		Clinical opinion
PNF	0		Clinical opinion
LFa	1		Clinical opinion

Resource use and unit cost estimation

Costing was conducted for the financial year 2012/13 and unit costs were acquired where possible from national sources including NHS reference costs,104 Unit Costs of Health and Social Care105 and British National Formulary (BNF). 106

Health measurement and valuation

The systematic review of economic evidence revealed that no preference-weighted quality-of-life values have been reported from patients with DC. However, a recently published study by Gu et al. elicited health-state utilities for DC from the general population using a discrete choice experiment (DCE). 107 At present, this is the only study that reports preference-weighted quality-of-life values and we incorporate these values into the model. In terms of the validity of the methods employed, this study adopted best practice to identify appropriate attributes and levels, via a focus group conducted in Scotland, and to employ appropriate, standard methods of design (orthogonal DCE fractional factorial design) and data analysis (conditional logistic modelling). Tests of rationality were also employed. The expected association between greater problems with contracture and reduction in health-state values was observed. There is no evidence, therefore, to suggest that the values are biased in any way through use of inappropriate methods.

In this study, health-state preferences were elicited via an internet survey from 1745 respondents resident in the UK. This study sample appeared to be relatively representative of the UK population. Participants were shown 10 pairs of hand profiles with both hands affected by DC and were required to identify a preferable hand. The eight joints (PIP and MCP joints of the index, middle, ring and little fingers) were each categorised for three different degrees of contracture (0°, 45°, 90°) and correspond to the Tubiana scale. The Tubiana scale is a staging system for DC which measures total flexion deformity for the PIP and MCP joints. The authors employed an anchoring method to restrict participants’ preferences onto the traditional 0–1 utility scale. Hands that were unaffected by DC were assigned a perfect utility score of 1, whereas the utility value for a hand that exhibited the worst possible DC (i.e. 90° of contracture in all eight joints) was derived by asking participants what levels of the European Quality of Life-5 Dimensions (EQ-5D) 5 levels of severity (EQ-5D-5L) profile would be most likely to be affected by living with this hand. Conditional logistic logit models were employed to estimate indirect utility estimates, which were subsequently rescaled to the anchor points on the EQ-5D-5L. The utility estimates were therefore dependent on degree of contracture, joint type and the finger in which the contracture occurred.

The van Rijssen et al. 70 and CORD I55 studies did not report which finger was affected by DC. Although mean baseline contracture and mean post-treatment contracture were reported in van Rijssen et al. ,70 CORD I55 reported the mean baseline total contracture index (this is the sum of fixed-flexion contractures in all 16 joints measured at screening). Both studies reported a mean of three joints affected per patient. We calculated the utility estimate from information on baseline patient characteristics in both studies. In both studies, this resulted in the average patient beginning at Tubiana stage 3 prior to treatment (> 90° but < 135°contracture of PIP and MCP joints). As the definition of treatment successes is to have a correction of contracture to within 0–5° of full extension, we assumed treatment successes would move from stage 3 to stage 0, whereas treatment failures would remain at Tubiana stage 3. Patients experiencing recurrence would move from stage 0 to stage 1 (45° or less contracture, but greater than 0); however, patients who received further treatment were assumed to move to stage 3 in the cycle prior to receiving treatment. This was to reflect that patients who proceed for further treatment following a recurrence are most likely to be clinically worse than patients who do not undergo any further treatment. We conducted a sensitivity analysis to assess the impact of patients who were clinically defined as a treatment failure achieving some quality-of-life improvement and, therefore, moving from stage 3 to stage 2.

To calculate health-state values, we calculated the pre- and post-treatment utility estimates for each finger individually and then applied a weighted estimate of mean utility at baseline and post treatment. The pre- and post-treatment utility values were estimated from the equation specified by Gu et al. :107

where U(hand_i) = the pre- or post-treatment value for hand_i, V (hand_i) = indirect utility weight for hand_i from the Conditional Logistic model estimate, V(best) = indirect utility weight for a hand with no contracture problems, V (anchor) = indirect utility weight associated with the worst (anchor) hand and EQ-5D_mean = the mean value of the EQ-5D-5L in the worst hand. To take a worked example of how a value is derived, consider the baseline patient, with Tubiana stage 3. First, we take the coefficient estimates reported in the study by Gu et al. 107 (supplementary article data, appendix B1) and reproduced below (Table 20). Then, to account for the mean patient at baseline, who has three joints affected, the indirect utility weights for Tubiana stage 3 are multiplied by 3 to give V baseline (with values for 1 and 2 used for the 1 joint and 2 joint models respectively, reported in the sensitivity analyses). U baseline is then derived from the formula in Equation 2, where V(best) = 0, V (anchor) = –23.5298 and EQ-5D mean = 0.4909.

This weighted utility estimate of 0.776268113 is then derived by multiplying the four calculated utility baseline estimates reported above by the probability of that finger being affected. Dias and Braybrooke29 reported the patterns of disease in the form of digit affected based on a large audit of the outcomes of surgery for DC, thereby enabling a calculation of this probability (Table 21). Values for Tubiana stages 0, 1 and 2 are calculated following the same procedure as outlined in the worked example. Table 22 presents the health-state utility values for each Tubiana stage as incorporated in the model.

Utility values for treatment complications

The clinician from our team identified major adverse events that would have an impact on quality of life (Mr David Lawrie, personal communication). No data on the utilities from adverse events following treatment for DC have been published. Thus, we assumed that patients experiencing a complication would receive a utility decrement of –0.0615; this is half a decrement on the European Quality of Life-5 Dimensions-3 levels (of severity) instrument (EQ-5D-3L) of a move from no pain or discomfort to some pain or discomfort. The utility decrement was applied for a time period of 6 weeks, which is the median referral to treatment waiting time for trauma and orthopaedics. After the 6 weeks it was assumed that patients would return to the appropriate utility value.

Comparison with previously published utility estimates

Utility estimates from the Gu et al. 107 study were included in this model as it was conducted in the UK and included a total of 1745 participants. Furthermore, the values are scaled in such a way that they have been estimated relative to an EQ-5D state. In comparison, the only other health-state valuation was reported by Chen et al. ,100 who constructed a utility survey specific to their study in a US setting and surveyed a total of 50 participants. They reported much higher utility values, ranging from 0.971 to 0.994, than those included in our model. The high utility values could be attributable to the standard gamble method of utility elicitation. This approach requires participants to choose between two alternatives. The first alternative is a treatment with two outcomes: (1) the patient is cured or; (2) the patient dies. Chen et al. 100 offered participants, in a hypothetical exercise, the opportunity to take a pill, which would lead to an immediate return to full health or death. The second alternative is to live in a particular chronic state over a number of years. In this case, participants could live the rest of their life with DC. However, as DC is not a life-threatening disease, participants may have been averse to choosing an alternative with a sudden risk of death, and this may explain the high utility values.

Discount rate (costs and benefits)

Costs and benefits (QALYs) were discounted at the NICE recommended rate of 3.5% per annum. We assessed the impact of discounting costs and benefits at a rate of 0% and 6% and also of discounting benefits at a rate of 1.5% while maintaining a discount rate of 3.5% for costs.

Time horizon

The analysis proceeded over a 37-year (i.e. lifetime) time horizon, following the cohort from age 63 to 100 years.

Deterministic sensitivity analysis

The process of populating the model required a number of parameter and structural assumptions. To assess the sensitivity of the base-case results to these assumptions, several deterministic sensitivity analyses were conducted. Sensitivity analyses were conducted on the success rates of treatment strategies, varying utility values, varying the cost of treatments, assuming patients had a lower number of mean joints affected at baseline and altering the discount rate. Where a strategy is dominated, either extendedly or absolutely, the reporting ICER is against the next less costly, non-dominated strategy. As PNF is the least costly option in all scenarios presented in the deterministic analysis, it automatically becomes the comparator. However, as PNF may not be available everywhere, the ICER is also reported separately for strategies that are extendedly dominated. The following is the list of the deterministic analyses that were undertaken:

1. number of injections to achieve a clinical success reduced to 1 for CCH

2. applied a 100% success rate to CCH

3. applied an equivalent success rate for CCH and LF (71%)

4. upper limit of the CI for the success rate for collagenase, with all other parameters the same as the base case

5. varied the success rate of CCH, holding all other parameters constant, so that the cost per QALY was under the threshold of £20,000 (79%)

6. applied a constant lifetime recurrence rate

7. lower limit of the CI for the recurrence rate for collagenase, with all other parameters the same as the base case

8. all treatments have the same recurrence rate, with all other parameters the same as the base case

9. both the upper limit of the CI for the success rate for collagenase and equal recurrence rates for all three treatments

10. assumed that all patients would proceed for further treatment following a recurrence

11. further treatment options for second-line PNF based on the van Rijssen trial (21% receive LF; 79% receive PNF)

12. CCH is not an option for second-line treatment after PNF (50% receive LF; 50% receive PNF)

13. applied a utility decrement for half a model cycle, post LF, to capture the potential recovery time following a surgical procedure

14. applied a baseline utility decrement for patients undergoing LF to assess the potential that patients may be feeling anxious prior to receiving a surgical procedure

15. applied a stage-2 utility value for treatment failures to reflect that even though patients may not have achieved clinical success, they may have gained some improvement in quality of life

16. applied a lower cost estimate for treatment strategies (CCH: £4584; PNF: £231; and LF: £1775)

17. applied a higher cost estimate for treatment strategies (CCH: £5011; PNF: £277; and LF: £2670)

18. assumed patients had a mean of 1 joint affected, with adjusted utility values to reflect the earlier success in the cycle as well as improved baseline utility

19. assumed patients had a mean of 1 joint affected, with adjusted utility values to reflect the earlier success and increased the recurrence rate for PNF to be equivalent to CCH

20. assumed patients had a mean of 1 joint affected, with adjusted utility values to reflect the earlier success with an equal proportion of patients receiving further treatment for PNF and CCH

21. assumed patients had a mean of 1 joint affected, with adjusted utility values to reflect the earlier success with an equal proportion of patients receiving further treatment for PNF and CCH and increased the recurrence rate for PNF to be equivalent to CCH

22. assumed patients had a mean of 2 joints affected, with adjusted utility values to reflect the earlier success in the cycle as well as improve baseline utility

23. assumed patients had a mean of 2 joints affected, with adjusted utility values to reflect the earlier success and increased the recurrence rate for PNF to be equivalent to CCH

24. assumed patients had a mean of 2 joints affected, with adjusted utility values to reflect the earlier success with an equal proportion of patients receiving further treatment for PNF and CCH

25. assumed patients had a mean of 2 joints affected, with adjusted utility values to reflect the earlier success with an equal proportion of patients receiving further treatment for PNF and CCH and increased the recurrence rate for PNF to be equivalent to CCH

26. discounting benefits at a rate of 1.5% per annum while maintaining a discount rate of 3.5% for costs

27. no discounting for costs and benefits

28. discounting costs and benefits at an annual rate of 6%.

Probabilistic sensitivity analysis

In order to characterise the uncertainty surrounding the selection of the optimal strategy, the model was analysed probabilistically. An appropriate distribution was assigned to hospital-based procedure unit costs, recurrence rates, treatment success rates, complication rates, probability of further treatment and utility values to reflect the uncertainty surrounding them. The probabilistic analysis proceeded by randomly selecting a value from the assigned distribution for the model parameters and recomputing the model results. This process was repeated 1000 times. To illustrate the uncertainty surrounding the estimates of cost-effectiveness, cost-effectiveness acceptability curves (CEACs) were generated using these 1000 estimates, using the NMB approach. CEACs demonstrate the probability of an intervention being cost-effective at different ceiling ratios of decision-makers’ WTP per QALY.

Mean costs, mean effects and incremental analysis

Table 23 presents the mean costs, mean QALYs and incremental cost per QALY gained associated with each strategy. Figure 33 presents the findings of cost per QALY gained graphically on the cost-effectiveness frontier. Strategies that do not fall on the line (the cost-effectiveness frontier) are strategies that are absolutely dominated (more costly and less effective than other strategies) and, therefore, they do not have the potential to be considered cost-effective. Strategies that fall on the line represent strategies that have the potential to be cost-effective, dependent on decision-makers’ WTP per QALY gained.

FIGURE 33.

Cost-effectiveness frontier based on cost per QALY.

The base-case results show PNF to be the least costly option, followed by LF and CCH. LF produced an increase in QALYs compared with both PNF and CCH. CCH was dominated by LF as it was more costly and less effective. The ICER for LF versus PNF was £10,871 per QALY gained (LF was more costly and more effective than PNF). Therefore, applying a ceiling WTP threshold of £20,000, LF was the favourable option from a cost-effectiveness perspective.

Deterministic sensitivity analysis

Table 24 shows the cost per QALY findings to the deterministic sensitivity analysis alterations to the findings of the base-case analysis. These analyses demonstrate that CCH has the most favourable cost-effectiveness only when it can achieve a success rate of at least 77%. There are some scenarios to which the cost per QALY estimate is sensitive for PNF. PNF becomes the preferred option in terms of cost-effectiveness when all patients proceed for further treatment and when the mean number of joints affected was assumed to be 1.

In the majority of scenarios presented in the deterministic analysis, LF remains the preferred option from a cost-effectiveness perspective.

Probabilistic sensitivity analysis

To assess the impact of combined uncertainty surrounding all model parameters and inputs, appropriate distributions were assigned and randomly sampled for each 1000 iterations of the base-case analysis. Owing to the limited data available, distributions were applied to hospital-based procedure unit costs, recurrence rates, treatment success rates, complication rates, probability of further treatment and utility values. These results were used to estimate the probability of each treatment strategy being preferred in terms of cost-effectiveness for different value decision-makers’ WTP for a QALY (Table 25). The resultant CEAC is illustrated in Figure 34. Figure 35 and Table 24 show that when applying a WTP threshold of £10,000 per QALY, PNF has a 53.2% probability of being cost-effective. However, at a WTP threshold of ≈ £14,000, the CEACs for PNF and LF cross. Implementing a WTP threshold of £20,000 per QALY, LF has a 64.3% probability of being considered cost-effective, followed by PNF (35.5%) and CCH (0.2%). Figure 35 shows the empirical estimate of the joint distribution of incremental costs and effects for PNF, CCH and LF. This figure clearly illustrates that CCH is dominated, in that it is more costly and less effective than LF, and PNF is the preferred option up until decision-makers’ WTP per QALY gained exceeds ≈ £14,000.

FIGURE 34.

Cost-effectiveness acceptability curve.

FIGURE 35.

Incremental cost-effectiveness scatterplot.

Differential results for subgroups according to patients with moderate and severe disease

Although few data were available from the identified studies to conduct subgroup analyses, the manufacturer’s submission presented the CCH treatment success rates and the mean number of joints affected for both patients with moderate and severe disease. Thus, two additional subgroup analyses, according to the severity of disease, were performed:

1. Population of patients with moderate disease and up to two affected joints. Patients with moderate disease were assumed to have a mean of 1.47 joints affected, with a mean of 1.6 CCH injections required per joint. CCH had a success rate of 80.7%. The cost of initial CCH treatment was £2359 (consisting of treatment of 1.47 joints), while the cost of CCH treatment per joint recurrence was £1605.

2. Population of patients with severe disease and up to two affected joints. Patients with severe disease were assumed to have a mean of 1.43 affected joints, with a mean of 1.6 CCH injections required per joint. The success rate of CCH was 53.7%. The cost of initial CCH treatment was £2295 (consisting of treatment of 1.43 joints), whereas the cost of CCH treatment per joint recurrence was £1605.

As fewer joints than the original base case were affected, utility values were revised. Expert opinion was used to identify five possible configurations of contracture for moderate and severe disease to derive appropriate utility values (Tables 26 and 27). For each configuration, a utility weight was calculated using the equations by Gu et al. for the dominant and non-dominant hands. An assumption was made as to the proportion of patients suffering Dupuytren’s disease in the dominant (67%) and non-dominant hand (33%). 107

A series of deterministic sensitivity analyses were conducted to test the sensitivity of the base-case results to the changes in key parameter values. Results of the additional analyses are presented separately for the moderate disease (Table 28) and the severe disease (Table 29) subgroups.

1. Base case: a mean of 1.47 affected joints; 1.6 CCH injections required; success rate of 80.7% for CCH; cost of initial CCH treatment £2359 and recurrent CCH treatment £1605.

2. The upper limit of the 95% CI surrounding the success rate for CCH (82.1%).

3. Constant lifetime recurrence, with recurrence rates the same as the original base case.

4. A 6-month probability of recurrence of 0.061 for all three treatments (based on the CORDLESS 5-year results) with recurrence occurring only in the first 5 years of the model.

5. Utility values calculated using the Assessment Group’s original method (i.e. condition specific with a utility of 1 for treatment success). Utility values: baseline, treatment failure = 0.950; treatment success = 1; recurrence = 0.965.

A probabilistic sensitivity analysis was conducted. A gamma distribution was assigned to hospital-based procedure unit costs and beta distributions were applied to recurrence rates, treatment success rates, complication rates, probability of further treatment and utility values to reflect the uncertainty surrounding them.

Figure 36 shows the CEAC for patients with moderate disease and Figure 37 shows the incremental cost-effectiveness scatterplot for patients with moderate disease. Figure 37 illustrates that, implementing a WTP threshold of £20,000 per QALY gained, PNF has a 40.6% probability of being cost-effective, followed by LF (39.7%) and CCH (19.7%).

FIGURE 36.

Cost-effectiveness acceptability curve: moderate disease.

FIGURE 37.

Incremental cost-effectiveness scatterplot: moderate disease.

Severe disease

Table 29 shows the same types of cost-effectiveness analyses presented in Table 28 for a population of patients with severe disease and up to two affected joints.

1. Base case: 1.43 affected joints; 1.6 injections of CCH required; 53.7% success rate of CCH; initial CCH treatment £2,359 and £1,605 for recurrent CCH treatment.

2. The upper limit of the 95% CI surrounding the success rate for CCH (82.1%).

3. Constant lifetime recurrence, with recurrence rates the same as the original base case.

4. A 6-month probability of recurrence of 0.061 for all three treatments (based on the CORDLESS 5-year results) with recurrence occurring only in the first 5 years of the model.

5. Utility values calculated using the Assessment Group’s original method (i.e. condition-specific with a utility of 1 for treatment success). Utility values: baseline and treatment failure = 0.925; treatment success = 1; and recurrence 0.965.

A probabilistic sensitivity analysis was also conducted for the severe group. A gamma distribution was assigned to hospital-based procedure unit costs and beta distributions were applied to recurrence rates, treatment success rates, complication rates, probability of further treatment and utility values to reflect the uncertainty surrounding them.

Figure 38 shows the CEAC for patients with severe disease and Figure 39 shows the incremental cost-effectiveness scatterplot for patients with severe disease. Figure 39 shows that, applying a WTP ratio of £20,000 per QALY gained, LF has a 45.1% probability of being considered cost-effective, followed by PNF (30.8%) and CCH (24.1%).

FIGURE 38.

Cost-effectiveness acceptability curve: severe disease.

FIGURE 39.

Incremental cost-effectiveness scatterplot: severe disease.

Summary of key results

The findings of the economic modelling suggest that, for patients with two or three joints affected, when considering QALYs as the unit of outcome, LF is likely to be the most cost-effective in comparison with PNF and CCH. The base-case estimate for LF, relative to PNF, is £10,871 per QALY gained. LF costs less and produces more QALYs than CCH. Based on the probabilistic sensitivity analysis, LF had a 64.3% chance of being cost-effective at a ceiling WTP ratio of £20,000 per QALY.

These estimates were generally found to be robust to uncertainty surrounding various model parameter inputs and assumptions, as illustrated by the deterministic sensitivity analysis.

The base-case finding that LF is the preferred option is driven by a number of factors: (1) greater QALY gains versus PNF and CCH; (2) lower cost compared with CCH; (3) lower clinical recurrence rate compared with PNF and CCH; and (4) higher treatment success rate compared with PNF and CCH.

Generalisability of results

Given that there are currently no head-to-head RCTs evaluating the strategies of interest, the effectiveness parameters within the model are largely populated by two studies from the USA and the Netherlands. This leads to a potential problem of transferability of effectiveness results to the UK, as surgeon quality and thresholds for surgery may not be the same in the UK. Furthermore, it is difficult to draw a definitive conclusion when the modelled patient population is not homogenous.

Attempts were made to identify EQ-5D data from UK cohorts for patients suffering from DC, but no data were identified. The modelling therefore relied on utility estimates derived from a recently published DCE study (deviation in part from the NICE reference case). At present, this is the only study that reports preference-weighted quality-of-life values in a UK setting.

Cost estimates were based on resource use in the UK according to expert opinion and national average unit costs were applied to resource-use estimates where possible.

Strengths and limitations

Attempts have been made to use the best available evidence for the economic model. The model does provide a flexible framework that can be easily updated to incorporate new evidence as it becomes available. This study has also identified gaps in the current literature and can identify the most important parameters that should be included in any future trials.

A major limitation is that the economic model is built from a naive indirect comparison and, as such, represents a departure from the NICE reference case. 102 The modelling was hampered by a dearth of suitable data and, as a result, a number of assumptions are inherent within the model. At present, there are no head-to-head RCTs or high-quality comparative studies comparing the relevant treatment strategies. This can be problematic as it introduces heterogeneity into the model if the patients’ characteristics differ between treatment arms of the individual studies. Moreover, no studies with sufficient follow-up to track patients post DC treatment were available. Therefore, uncertainty remains regarding the applicability of many parameter model inputs. This emphasises the importance of further research to be conducted in the UK.

Conclusions

Overall, the naive indirect comparison, the level of uncertainty surrounding the model inputs and the large number of assumptions used makes it challenging to draw definitive conclusions of the cost-effectiveness of CCH, PNF and LF for the treatment of adults with DC. However, our modelling does suggest that if success and recurrence rates included in the model are reflective of those in the UK general population, then LF is likely to offer the most cost-effective approach, for patients with two or three joints affected. The subgroup probabilistic sensitivity analyses indicate that for patients with moderate disease and fewer joints affected, PNF is the preferred treatment option, whereas for patients with severe disease and fewer joints affected, LF has the highest probability of being cost-effective (45.1%). Future research should comprise a well-designed RCT, which would enable the generation of comparable estimates, in the form of recurrence rates, success and failure rates, complications and a measurement of utility from the EQ-5D, across the treatment strategies of interest.

Principal findings: clinical effectiveness

No head-to-head randomised trials of collagenase versus surgery were identified in the current literature. For the purpose of this assessment we considered evidence from:

• five RCTs comparing collagenase with placebo (493 participants)

• fifteen case series assessing the effects and safety of collagenase (3154 participants)

• two non-randomised comparative studies of collagenase versus surgery (105 participants)

• three RCTs comparing different surgical procedures (334 participants) and

• five non-randomised comparative studies assessing different surgical procedures (3571 participants).

Of the five RCTs comparing collagenase with placebo, three55,56,65 provided outcome measures that could be assessed in meta-analyses. Only one trial complied with the time horizon of 12 months’ follow-up recommended by the European Medicines Agency for efficacy trials, whereas both the CORD I and CORD II trials had a double-blind, randomised phase of 90 days and a subsequent open-label phase of 9 months.

• Primary MCP joints and PIP joints treated with collagenase were significantly more likely to achieve clinical success (i.e. reduction of contracture to 0–5° of normal 30 days after last injection) or clinical improvement (i.e. reduction in contracture of at least 50% 30 days after last injection) than those treated with placebo, with a greater reduction observed in MCP joints than in PIP joints.

• Significantly more participants treated with collagenase experienced adverse events (i.e. peripheral oedema, contusion, pain in extremity, injection site pain, injection site haemorrhage, injection site swelling, tenderness, pruritus, lymphadenopathy, axillary pain), with the exception of injection site vesicles, where the risk was low and similar for both treatment groups. In addition, four serious adverse events were reported among participants treated with collagenase (one case of complex regional pain syndrome and two cases of tendon rupture in CORD I55 and one case of flexion pulley rupture in CORD II56). It is worth also noting that two collagenase case series79,85 reported a total of four serious adverse events after collagenase injections: two flexor tendon ruptures, one pulley rupture and one episode of pain requiring hospitalisation.

• No recurrences were observed in the CORD I55 and CORD II56 trials during the study period (90 days). Badalamente and Hurst5 reported that five joints (four PIP joints and one MCP joint) had a recurrence over the 24-month follow-up period (12-month controlled trial followed by an open-label phase of 12 months).

Rates of recurrence for collagenase were derived from the Badalamente et al. RCT64 and single-arm observational studies (collagenase case series):

• In general, recurrence rates for MCP joints varied from 0% at 1 year to 27% at 3 years, whereas those for PIP joints varied from 0% at 1 year to 56% at 3 years.

• The manufacturer provided 5-year data for the CORDLESS study that included patients from previous collagenase cohorts (e.g. CORD I, CORD II, JOINT I, JOINT II). The rate of recurrence for successfully treated joints at 5 years was 46.7%.

• One small study reported the results of eight patients treated with collagenase (six treated for MCP joint contracture and two for PIP joint contracture) who completed the 8-year assessment, the longest follow-up reported in the literature. The observed recurrence rate was 67% (4/6) for MCP joints and 100% (2/2) for PIP joints.

The two small, non-randomised studies, at high risk of bias, that compared collagenase with fasciectomy71 and PNF72 showed that:

• MCP joints treated with collagenase had a similar reduction in contracture from baseline than those treated with fasciectomy (mean post treatment: 3.6° vs. 3.7°). However, PIP joints treated with fasciectomy had a greater reduction in contracture (mean post treatment: 8.1°) from baseline than those treated with collagenase (mean post treatment: 17.5°). 71

• A greater proportion of both MCP and PIP joints treated with PNF showed reduction in contracture (67%) than collagenase-treated joints (56%). 72

• No recurrences were observed during the study period (mean follow-up was 6 months in one study and 32–39 months in the other study).

Owing to the small sample size and the high risk of bias of these two non-randomised comparative studies, it is not possible to draw any firm conclusions on the relative efficacy of collagenase versus either fasciectomy or PNF.

The RCTs and non-randomised comparative studies assessing different surgical techniques tended towards greater clinical success for MCP joints than PIP joints, with slightly higher success rates for fasciectomy than fasciotomy.

Randomised controlled trials assessing different surgical interventions reported rates of recurrence that ranged from 13% for fasciectomy at 3 years to 85% for PNF at 5 years. It is worth noting that the 85% recurrence rate was reported by the van Rijssen et al. study70 that assessed the effects and complications of PNF compared with fasciectomy. van Rijssen et al. defined recurrence as an increase of total passive extension deficit of at least 30° compared with the 6-week post-operative values in all treated hands. 70 However, if the definition of the CORD trials was applied (i.e. return of contracture of at least 20° in successfully treated joints), the 5-year recurrence rate for PNF would be 21.8% for MCP joints and 23.5% for PIP joints (see Table 6), considerably lower than that reported for collagenase. Recurrence rates in non-randomised comparative studies ranged from 0% to 50% for fasciectomy at around 3 years and from 15% to 50% for open fasciotomy at around 2 years. Mild adverse events such as oedema, ecchymosis, skin tears, swelling, infection, haematoma, pain and delayed wound healing were reported more often after fasciectomy than PNF, fasciotomy or dermofasciectomy.

The inconsistencies in the definitions and measurement of treatment efficacy, characteristics of participants, and choice of comparators across studies made any comparison of data challenging. Given that no common comparator treatment was identified between the various comparative studies, it was impossible to conduct an indirect meta-analysis as recommended by the NICE methods guide.

In brief, collagenase proved to be superior to placebo in short-term RCTs, with better results for MCP than PIP joints. The long-term effects of collagenase are yet to be determined. Similarly, the treatment of recurrences after collagenase injections needs to be clarified. It is presently unclear whether collagenase can be used as a second-line treatment for DC. Further surgery (i.e. LF) after collagenase could be potentially more difficult than standard surgical revision98 (Mr David Lawrie, personal communication). Observational data indicated a recurrence rate of 47% at 5 years, which is not dissimilar to, and even higher than, the recurrence rates of LF and PNF reported in some cohorts in the literature. However, collagenase has not yet been compared with any surgical interventions in RCTs or high-quality comparative studies.

Principal findings: cost-effectiveness

The manufacturer submitted an economic evaluation, which suggested that collagenase was cost-effective. However, our economic modelling found that under base-case assumptions, LF is the preferable option from a cost-effectiveness perspective. This estimate was generally found to be robust to varying key model parameter inputs and assumptions. These findings stem from the higher success rate and lower recurrence rate compared with collagenase and PNF. However, the deterministic sensitivity analyses show that PNF may be a cost-effective alternative to LF when the proportion of patients proceeding for further treatment are varied. PNF is the most cost-effective option when all patients proceeded for further treatment following a recurrence. The modelling suggests that the success rate of collagenase would need to increase from 63% at base case to 77% if it were to offer a cost-effective alternative to PNF or LF. This is due to the higher initial treatment cost and higher recurrence rate. Treatment costs can, however, be reduced through patient selection (i.e. where one joint is affected, costs are lower compared with LF, although this alone would not be sufficient to make collagenase less costly than PNF and so it remains dominated in this analysis).

A probabilistic sensitivity analysis was conducted to assess the impact of combined uncertainty surrounding certain model parameters and inputs. Owing to the limited data available, distributions were applied to hospital-based procedure unit costs, recurrence rates, treatment success rates, complication rates, probability of further treatment and utility values and were randomly sampled for 1000 iterations of the base-case analysis. The probabilistic sensitivity analysis suggest that at a WTP for a QALY gained of £20,000, the chance of collagenase, PNF and LF being the most cost-effective treatment strategy is 0.2%, 35.5% and 64.3%, respectively. The case for cost-effectiveness of LF increases to 71.4% and 72.1% at threshold values of WTP of £30,000 and £50,000.

The subgroup analyses indicate that collagenase is dominated in the majority of scenarios and did not produce an ICER under the threshold of £20,000 per QALY. The results from the probabilistic sensitivity analyses show that for patients with moderate disease and with fewer joints affected, PNF may offer a cost-effective treatment option, whereas for patients with severe disease, LF has the highest probability of being cost-effective.

Limitations of the review of clinical effectiveness

The main limitation of this assessment was the lack of comparative evidence on collagenase versus surgery and the small evidence base for estimating the effects of specific surgical procedures (i.e. fasciectomy and PNF). In particular, the review of clinical effectiveness was hampered by (1) the lack of randomised trials or high-quality comparative studies assessing collagenase versus surgery; (2) the use of observational studies for estimating rates of recurrence and complications after collagenase treatment; and (3) the major inconsistencies in the way efficacy was measured and reported across studies.

The substantial differences across studies in terms of characteristics of patient population (e.g. age, sex, severity of disease), measures of efficacy (e.g. changes in degree contracture, passive or active extension deficit, total passive extension deficit, percentage of improvement), population denominator (e.g. by patient, by hand, by finger, by joint), definition of recurrence, and length of follow-up inevitably limited the comparability of evidence.

Strengths and limitations of the economic evaluation

Attempts were made to use the best available evidence to populate the economic model. The model provided a flexible framework that can be updated to incorporate new evidence as it becomes available. This study has also highlighted gaps in the current literature and identified the most important parameters that should be included in any future trials. The conduct of a wide range of sensitivity analysis helped to offset some of the uncertainty and to identify the most important drivers of cost-effectiveness.

There were a number of limitations associated with the estimates used to populate the model parameters, and the results should be interpreted with caution. A major limitation was that the economic model was built from a naive indirect comparison of data and, as such, represents a departure from the NICE reference case. The modelling was hampered by a dearth of suitable data and, as a result, many untestable assumptions are built within the model. At present, there are no RCTs or high-quality comparative studies comparing the relevant treatment strategies. This can be problematic as it introduces heterogeneity into the model if the patients’ characteristics differ between treatment arms of the individual studies. Moreover, no studies with sufficient follow-up to track patients post DC treatment were available. Therefore, uncertainty remains regarding the appropriateness of many parameter model inputs. This emphasises the importance of further research to be conducted in the UK.

Clinical effectiveness

The comparative efficacy of collagenase versus any surgical procedures has yet to be demonstrated. Evidence on long-term effectiveness, recurrence rate and risk of complications is still lacking. Second-line treatments after unsuccessful collagenase injections are yet to be defined.

There was considerable variation in measurements of contractures, population denominators and definitions of success and recurrence across included studies.

At present, there is little consistency in the way outcomes for Dupuytren’s disease are measured and reported. In particular, there are no standard methods or consensus for assessing the degree of contracture, defining recurrence and determining the length of follow-up. Moreover, indications for specific surgical procedures (e.g. PNF) and for second-line treatments after disease recurrence (e.g. collagenase, PNF) are not clearly defined.

Cost-effectiveness

There was substantial uncertainty surrounding the values for many of the variables in the model, and so the estimated ICERs should be interpreted with caution. Estimates of utilities for health states in the model were indirectly derived from a recently published DCE rather than directly measured from a preference-weighted quality-of-life instrument. Thus, the extent to which changes in quality of life have been adequately captured, in the sense of them representing valid estimates of change on the usual 0 = dead, 1 = full health scale, is unknown. Direct measurement of health-related quality of life using a generic preference-based instrument, such as the EQ-5D, the Short Form questionnaire-12 items or the Health Utilities Index-3, would more accurately capture the impact of improvement in joint contracture that is offered by all treatments. Although the deterministic and probabilistic analyses address some of these uncertainties, there is an underlying weakness in the clinical effectiveness evidence base that cannot be dealt with through sensitivity analysis, no matter how extensively this is conducted.

Overall, the naive indirect comparison, the level of uncertainty surrounding the model inputs and the large number of assumptions used makes it challenging to draw definitive conclusions on the cost-effectiveness of collagenase relative to PNF and LF for the treatment of adults with DC. However, our modelling does suggest that if the parameter estimates are broadly reflective of those likely to apply in UK clinical practice, then LF is likely to offer the most cost-effective approach for DC patients with more than one joint affected.

EMBASE

Date searched from: 1990 to 2014 Week 8.

Ovid MEDLINE

Date searched from: 1990 to February Week 2 2014.

Ovid MEDLINE In-Process & Other Non-Indexed Citations

Date searched from: 25 February 2014.

OVID multifile search URL: https://shibboleth.ovid.com/

1. Dupuytren Contracture/

2. dupuytren$.tw.

3. (palm$ adj3 fibromatosis).tw.

4. or/1-3

5. Microbial Collagenase/ use mesz

6. clostridiopeptidase A/ use oemez

7. (collagenase or xiapex or xiaflex).tw.

8. or/5-7

9. Dupuytren Contracture/su

10. (fasciectom$ or fasciotom$).tw.

11. (dermatofasciectom$ or dermofasciectom$ or dermo fasciectom$).tw

12. aponeurotom$.tw.

13. ((open or closed) adj2 palm technique?).tw.

14. or/9-13

15. 4 and (8 or 14)

16. exp clinical trial/ use oemez

17. randomized controlled trial.pt.

18. controlled clinical trial.pt.

19. randomization/ use oemez

20. randomi?ed.ab.

21. placebo.ab.

22. drug therapy.fs.

23. randomly.ab.

24. trial.ab

25. groups.ab

26. or/16-25

27. comparative study/ use mesz

28. follow-up studies/ use mesz

29. time factors/ use mesz

30. Treatment outcome/ use oemez

31. major clinical study/ use oemez

32. controlled study/ use oemez

33. (chang$ or evaluat$ or reviewed or baseline).tw.

34. (prospective$ or retrospective$).tw.

35. (cohort$ or series).tw.

36. (compare$ or compara$).tw.

37. or/27-36

38. 15 and (26 or 37)

39. exp animals/ not humans/

40. nonhuman/ not human

41. 38 not (39 or 40)

42. case report/ use oemez

43. case reports.pt.

44. 41 not (42 or 43)

45. remove duplicates from 44

Science Citation Index

Date searched from: 1990–27 February 2014.

Bioscience Information Service (BIOSIS)

Date searched from: 1990–27 February 2014.

Conference Proceedings Citation Index – Science

Date searched from: 2012–27 February 2014.

ISI Web of Knowledge

URL: http://wok.mimas.ac.uk/

# 1 TS=dupuytren*

# 2 TS=(palm* NEAR/3 fibromatosis).

# 3 #1 OR #2

# 4 TS=(collagenase or xiapex ot xiaflex)

# 5 TS=(fasciectom* or fasciotom*).# 6 TS=(dermatofasciectom* or dermofasciectom* or dermo fasciectom*).

# 7 TS=aponeurotom*

# 8 TS= ((open or closed) NEAR/2 "palm technique*")

# 9 #8 OR #7 OR #6 OR #5 OR #4

#10 #9 AND #3

Centre for Reviews and Dissemination

URL: http://nhscrd.york.ac.uk/welcome.htm

1. MeSH DESCRIPTOR Dupuytren Contracture

2. (dupuytren)

3. #1 OR #2

Clinical Trials

Date searched: February 2014.

URL: http://clinicaltrials.gov/ct/gui/c/r

Topic= Dupuytren Contracture

International Clinical Trials Registry Platform (ICTRP)

Date searched: June 2013.

World Health Organization

URL: http://www.who.int/ictrp/en/

Dupuytren*

EMBASE

Date searched from: 1990 to 2014 Week 8.

Ovid MEDLINE(R)

Date searched from: 1990 to February Week 2 2014.

Ovid MEDLINE(R) In-Process & Other Non-Indexed Citations

Date searched: 25 February 2014.

OVID multifile search URL: https://shibboleth.ovid.com/

1. Dupuytren Contracture/

2. dupuytren$.tw.

3. (palm$ adj3 fibromatosis).tw

4. or/1-3

5. exp “costs and cost analysis”/ use mesz

6. exp economic evaluation/ use emez

7. economics/

8. health economics/ use emez

9. exp economics,hospital/ use mesz

10. exp economics,medical/ use mesz

11. economics,pharmaceutical/ use mesz

12. exp budgets/

13. exp models, economic/ use mesz

14. exp decision theory/

15. monte carlo method/

16. markov chains/

17. exp technology assessment, biomedical/

18. cost$.ti.

19. (cost$ adj2 (effective$ or utilit$ or benefit$ or minimis$)).ab.

20. economics model$.tw.

21. (economic$ or pharmacoeconomic$).tw.

22. (price or prices or pricing).tw.

23. (value adj1 money).tw.

24. markov$.tw.

25. monte carlo.tw.

26. (decision$ adj2 (tree? or analy$ or model$)).tw.

27. or/5-26

28. 4 and 27

29. remove duplicates from 28

Database: Health Management Information Consortium (HMIC)

Date searched from: 1979 to February 2014.

URL: https://auth.athensams.net/

1. dupuytrens contracture/

2. dupuytren$.tw.

3. (palm$ adj3 fibromatosis).tw.

4. 1 or 2 or 3

Science Citation Index

Date searched from: 1990–27 February 2014.

ISI Web of Knowledge

URL: http://wok.mimas.ac.uk/

# 1 TS=dupuytren*

# 2 TS=cost*

# 3 TS=(decision* N/2 (tree* or analy* or model))

# 4 TS=(markov or monte carlo)

# 5 TS=economic*

# 6 #5 OR #4 OR #3 OR #2

# 7 #6 AND #1

NHS Economic Evaluation Database (EED)

Seached from: February 2014.

Centre for Reviews and Dissemination

URL: http://nhscrd.york.ac.uk/welcome.htm

1. MeSH DESCRIPTOR Dupuytren Contracture

2. (dupuytren)

3. #1 OR #2

RePEc (Research Papers in Economics). URL: http://repec.org/

Dupuytren

EMBASE

Date searched from: 1990 to 2014 Week 8.

Ovid MEDLINE(R)

Date searched from: 1990 to February Week 2 2014.

Ovid MEDLINE In-Process & Other Non-Indexed Citations

Date seached: 25 February 2014.

OVID multifile search: URL: https://shibboleth.ovid.com/

1. Dupuytren Contracture/

2. dupuytren$.tw.

3. (palm$ adj3 fibromatosis).tw.

4. or/1-3

5. quality of life/

6. quality adjusted life year/

7. “Value of Life”/ use mesz

8. health status indicators/ use mesz

9. health status/ use emez

10. sickness impact profile/ use mesz

11. disability evaluation/ use mesz

12. disability/ use emez

13. activities of daily living/ use mesz

14. exp daily life activity/ use emez

15. cost utility analysis/ use emez

16. rating scale/

17. questionnaires/

18. (quality adj1 life).tw.

19. quality adjusted life.tw.

20. disability adjusted life.tw.

21. (qaly? or qald? or qale? or qtime? or daly?).tw

22. (euroqol or euro qol or eq5d or eq 5d).tw.

23. (hql or hqol or h qol or hrqol or hr qol).tw.

24. (hye or hyes).tw

25. health$ year$ equivalent$.tw.

26. (hui or hui1 or hui2 or hui3).tw

27. (health adj3 (utilit$ or disutili$)).tw.

28. (health adj3 (state or status)).tw.

29. (sf36 or sf 36 or short form 36 or shortform 36).tw

30. (sf6 or sf 6 or short form 6 or shortform 6).tw.

31. (sf12 or sf 12 or short form 12 or shortform 12).tw.

32. (sf16 or sf 16 or short form 16 or shortform 16).tw.

33. (sf20 or sf 20 or short form 20 or shortform 20).tw.

34. willingness to pay.tw.

35. standard gamble.tw

36. trade off.tw.

37. conjoint analys?s.tw.

38. discrete choice.tw.

39. or/5-38

40. (case report or editorial or letter).pt.

41. case report/

42. 4 and 39

43. 42 not (40 or 41)

44. remove duplicates from

Science Citation Index (1990–27 February 2014)

Cost-effectiveness Analysis Registry (CEA)

Date searched: February 2014.

URL https://research.tufts-nemc.org/cear4/default.asp

Dupuytren

Websites consulted

Agency for Healthcare Research and Quality, URL: www.ahrq.gov/

All Wales Medicine Strategy Group, URL: www.awmsg.org/

American Society for Surgery of the Hand, URL: www.assh.org/

Belgian Health Care Knowledge Centre (KCE), URL: https://kce.fgov.be/

British Dupuytren’s Society, URL: http://dupuytrens-society.org.uk/

Canadian Agency for Drugs and Technologies in Health, URL: www.cadth.ca/

European Medicine Agency, URL: www.ema.europa.eu/ema/

Federation of European Societies for Surgery of the Hand, URL: www.fessh2014.com/

French National Authority for Health (HAS), URL: www.has-sante.fr/

Health Information and Quality Authority, URL: www.hiqa.ie/

Institute for Clinical and Economic Review, URL: www.icer-review.org/

Institute for Quality and Efficiency in Health Care, URL: www.iqwig.de/

International Dupuytren Society, URL: www.dupuytren-online.info/

International Symposium on Dupuytren’s Disease, URL: http://dupuytrensymposium.com/

Medicines and Healthcare Products Regulatory Agency, URL: www.mhra.gov.uk/

Medical Services Advisory Committee, Australia, URL: www.msac.gov.au/

National Institute for Health and Care Excellence, URL: www.nice.org.uk/

NHS Quality Improvement Scotland, URL: www.healthcareimprovementscotland.org/

Scottish Medicines Consortium, URL: www.scottishmedicines.org.uk/

US Food and Drug Administration, URL: www.fda.gov/default.htm

Version 1, August 2012

Badalamente and Hurst48

• Badalamente MA, Hurst LC. Enzyme injection as nonsurgical treatment of Dupuytren’s disease. J Hand Surg Am 2000;25:629–36.

Badalamente et al.64

• Badalamente MA, Hurst LC, Hentz VR. Collagen as a clinical target: nonoperative treatment of Dupuytren’s disease. J Hand Surg Am 2002;27:788–98.

Badalamente and Hurst63

• Badalamente M, Hurst L. Nonsurgical treatment of Dupuytren’s disease. J Investig Med 2005;53:S394.

Badalamente and Hurst65

• Badalamente MA, Hurst LC. Efficacy and safety of injectable mixed collagenase subtypes in the treatment of Dupuytren’s contracture. J Hand Surg Am 2007;32:767–74.

Badalamente and Hurst77

• Badalamente MA, Hurst LC. Longer-term follow-up of recurrence after collagenase injection for Dupuytren’s contracture level 1 evidence. J Hand Surgery Am 2011;36(Suppl. 1):14.

Bainbridge et al.73

• Bainbridge C, Dahlin LB, Szczypa PP, Cappelleri JC, Guerin D, Gerber RA. Current trends in the surgical management of Dupuytren’s disease in Europe: an analysis of patient charts. Eur Orthop Traumatol 2012;3:31–41.

• Dias JJ, Singh HP, Ullah A, Bhowal B, Thompson JR. Patterns of recontracture after surgical correction of Dupuytren disease. J Hand Surg 2013;38:1987–93.

Citron and Hearnden74

• Citron N, Hearnden A. Skin tension in the aetiology of Dupuytren’s disease; a prospective trial. J Hand Surg Br 2003;28:528–30.

Citron and Nunez67

• Citron ND, Nunez V. Recurrence after surgery for Dupuytren’s disease: a randomized trial of two skin incisions. J Hand Surg Br 2005;30:563–6.

Coleman et al.78

• Coleman S, Gilpin D, Tursi J, Kaufman G, Jones N, Cohen B. Multiple concurrent collagenase clostridium histolyticum injections to Dupuytren’s cords: an exploratory study. BMC Musculoskelet Disord 2012;13:61.

Coleman et al.79

• Coleman S, Gilpin D, Kaplan FT, Houston A, Kaufman GJ, Cohen BM, et al. Efficacy and safety of concurrent collagenase clostridium histolyticum injections for multiple Dupuytren contractures. J Hand Surg Am 2014;39:57–64.

Considine and Hirpara80

• Considine S, Hirpara KM. Early results of the use of collagenase in the treatment of Dupuytren’s contracture. Ir J Med Sci 2013;182(Suppl. 2):52–3.

Gilpin et al.56

• Gilpin D, Coleman S, Hall S, Houston A, Karrasch J, Jones N. Injectable collagenase clostridium histolyticum: a new nonsurgical treatment for Dupuytren’s disease. J Hand Surg Am 2010;35:2027–38.

Hayton et al.81

• Hayton MJ, Bayat A, Chapman DS, Gerber RA, Szczypa PP. Isolated and spontaneous correction of proximal interphalangeal joint contractures in Dupuytren’s disease: an exploratory analysis of the efficacy and safety of collagenase clostridium histolyticum. Clin Drug Invest 2013;33:905–12.

Hurst et al.55

• Hurst LC, Badalamente MA, Hentz VR, Hotchkiss RN, Kaplan FTD, Meals RA, et al. Injectable collagenase clostridium histolyticum for Dupuytren’s contracture. N Engl J Med 2009;361:968–79.

• Witthaut J, Bushmakin AG, Gerber RA, Cappelleri JC, Le Graverand-Gastineau MP. Determining clinically important changes in range of motion in patients with Dupuytren’s Contracture: secondary analysis of the randomized, double-blind, placebo-controlled CORD I study. Clin Drug Invest 2011;31:791–8.

Kaplan et al.83

• Kaplan FTD, Badalamente M, Hurst L, Merrell GA, Pahk R. Delayed manipulation following clostridial collagenase histolyticum injection for dupuytren contracture. J Hand Surg Am 2013;38(Suppl. 10):e52–3.

Martin-Ferrero et al.84

• Martin-Ferrero MA, Simon-Perez C, Rodriguez-Mateos JI, Garcia-Medrano B, Hernandez-Ramajo R, Brotat-Garcia M. [Treatment of Dupuytren’s disease using collagenase from Clostridium histolyticum.] Rev Esp Cir Ortop Traumatol 2013;57:398–402.

McMahon et al.85

• McMahon HA, Bachoura A, Jacoby SM, Zelouf DS, Culp RW, Osterman AL. Examining the efficacy and maintenance of contracture correction after collagenase clostridium histolyticum treatment for Dupuytren’s disease. Hand 2013;8:261–6.

Naam71

• Naam NH. Functional outcome of collagenase injections compared with fasciectomy in treatment of Dupuytren’s contracture. Hand 2013;8:410–16.

Nydick et al.72

• Nydick JA, Olliff BW, Garcia MJ, Hess AV, Stone JD. A comparison of percutaneous needle fasciotomy and collagenase injection for Dupuytren disease. J Hand Surg Am 2013;38:2377–80.

Peimer et al.86

• Peimer CA, Skodny P, Mackowiak JI. Collagenase clostridium histolyticum for Dupuytren-contracture: patterns of use and effectiveness in clinical practice. J Hand Surg Am 2013;38:2370–6.

Peimer et al.87

• Peimer CA, Blazar P, Coleman S, Kaplan FT, Smith T, Tursi JP, et al. Dupuytren contracture recurrence following treatment with collagenase clostridium histolyticum (CORDLESS study): 3-year data. J Hand Surg Am 2013;38:12–22.

• Kaplan FTD, Hentz VR, Tursi J, Kaufman G, Smith T. Patterns of recurrent contracture in patients with Dupuytren’s disease 3 years following successful treatment with collagenase clostridium histolyticum: Level 4 evidence. J Hand Surg Am 2012;37(Suppl. 1):14.

Ribak et al.57

• Ribak S, Borkowski J, Amaral RP, Massato A, Ávila I, de Andrade D. Dupuytren contracture: comparative study between partial fasciotomy and percutaneous fasciotomy. J Braz Orthoped 2013;48:545–53.

Skirven et al.88

• Skirven TM, Bachoura A, Jacoby SM, Culp RW, Osterman AL. The effect of a therapy protocol for increasing correction of severely contracted proximal interphalangeal joints caused by dupuytren disease and treated with collagenase injection. J Hand Surg Am 2013;38:684–9.

Skoff75

• Skoff HD. The surgical treatment of Dupuytren’s contracture: a synthesis of techniques. Plast Reconstr Surg 2004;113:540–4.

Syed et al.91

• Syed A, Mcfarlane J, Chester T, Powers D, Sibly F, Talbot-Smith A. Clinical efficacy and cost-effectiveness of Clostridium histolyticum collagenase injections in a subpopulation of Dupuytren’s contracture patients. Eur Orthop Traumatol 2013;5:311–16.

Toppi et al.76

• Toppi JT, Trompf L, Smoll NR, Lim V, Smith K, Findlay MW, et al. Dupuytren’s contracture: an analysis of outcomes of percutaneous needle fasciotomy versus open fasciectomy [published online ahead of print]. ANZ J Surg 2014.

Ullah et al.68

• Ullah AS, Dias JJ, Bhowal B. Does a ‘firebreak’ full-thickness skin graft prevent recurrence after surgery for Dupuytren’s contracture?: a prospective, randomised trial. J Bone Joint Surg Br 2009;91:374–8.

van Rijssen et al.69,70

• van Rijssen AL, Gerbrandy FS, ter Linden H, Klip H, Werker PM. A comparison of the direct outcomes of percutaneous needle fasciotomy and limited fasciectomy for Dupuytren’s disease: a 6-week follow-up study. J Hand Surg Am 2006;31:717–25.

• van Rijssen AL, ter Linden H, Werker PM. Five-year results of a randomized clinical trial on treatment in Dupuytren’s disease: percutaneous needle fasciotomy versus limited fasciectomy. Plast Reconstr Surg 2012;129:469–77.

Watt et al.89

• Watt AJ, Curtin CM, Hentz VR. Collagenase injection as nonsurgical treatment of Dupuytren’s disease: 8-year follow-up. J Hand Surg Am 2010;35:534–9.

Witthaut et al.90

• Witthaut J, Jones G, Skrepnik N, Kushner H, Houston A, Lindau TR. Efficacy and safety of collagenase clostridium histolyticum injection for Dupuytren contracture: short-term results from 2 open-label studies. J Hand Surg Am 2013;38:2–11.

Ineligible study design (n = 34)

Anonymous. Collagenase clostridium histolyticum. Short-term benefit in Dupuytren’s contracture, unknown long-term risks. Prescrire Int 2011;20:285–7.

Bainbridge C, Gerber RA, Szczypa PP, Smith T, Kushner H, Cohen B, et al. Efficacy of collagenase in patients who did and did not have previous hand surgery for Dupuytren’s contracture. J Plast Surg Hand Surg 2012;46:177–83.

Ball C, Pratt AL, Nanchahal J. Optimal functional outcome measures for assessing treatment for Dupuytren’s disease: a systematic review and recommendations for future practice. BMC Musculoskel Disord 2013;14:131.

Boyer MI, Gelberman RH. Complications of the operative treatment of Dupuytren’s disease. Hand Clin 1999;15:161–6.

Breed CM, Smith PJ. A comparison of methods of treatment of pip joint contractures in Dupuytren’s disease. J Hand Surg Br 1996;21B:246–51.

Brenner P. Dupuytren’s disease of ring and little finger. Orthop Traumatol 2002;10:138–58.

Chen W, Zhou H, Pan ZJ, Chen JS, Wang L. The role of skin and subcutaneous tissues in Dupuytren’s contracture: an electron microscopic observation. Orthopaed Audio-Synopsis Contin Med Educ 2009;1:216–21.

Chick LR, Lister GD. Surgical alternatives in Dupuytren’s contracture. Hand Clin 1991;7:715–9.

Crean SM, Gerber RA, Le Graverand MP, Boyd DM, Cappelleri JC. The efficacy and safety of fasciectomy and fasciotomy for Dupuytren’s contracture in European patients: a structured review of published studies. J Hand Surg Eur 2011;36:396–407.

Denkler K. Surgical complications associated with fasciectomy for Dupuytren’s disease: a 20-year review of the English literature. Eplasty 2010;10:e15.

Denkler K. Collagenase for recurrent Dupuytren contracture with skin grafts. J Hand Surg Am 2013;38:1264.

Diaz R, Curtin C. Needle aponeurotomy for the treatment of Dupuytren’s disease. Hand Clin 2014;30:33–8.

Friedrich JB. Discussion: extensive percutaneous aponeurotomy and lipografting: a new treatment for Dupuytren disease. Plast Reconstr Surg 2011;128:229–30.

Hallock GG. Skin laceration as a serious adverse sequela of injectable collagenase for Dupuytren contracture. Plastic Reconstruct Surg 2012;129:205–6e.

Hentz VR. Collagenase injections for treatment of Dupuytren disease. Hand Clin 2014;30:25–32.

Hurst LC, Badalamente MA, Hentz VR, Hotchkiss RN, Kaplan FTD, Meals RA, et al. Injectable collagenase clostridium histolyticum for Dupuytren’s contracture. N Engl J Med 2009;361:968–79.

Jerosch-Herold C, Shepstone L, Chojnowski AJ, Larson D, Barrett E, Vaughan SP. Night-time splinting after fasciectomy or dermo-fasciectomy for Dupuytren’s contracture: a pragmatic, multi-centre, randomised controlled trial. BMC Musculoskel Dis 2011;12:136.

Kan HJ, Verrijp FW, Huisstede BMA, Hovius SER, Van Nieuwenhoven CA, Selles RW. The consequences of different definitions for recurrence of Dupuytren’s disease. J Plast Reconstruct Aest Surg 2013;66:95–103.

Ketchum LD. The use of the full thickness skin graft in Dupuytren’s contracture. Hand Clin 1991;7:731–41.

Knobloch K, Kuehn M, Vogt PM. Focused extracorporeal shockwave therapy in Dupuytren’s disease – a hypothesis. Med Hypotheses 2011;76:635–7.

Larocerie-Salgado J, Davidson J. Nonoperative treatment of PIPJ flexion contractures associated with Dupuytren’s disease. J Hand Surg Eur 2012;37:722–7.

Lubahn JD. Open-palm technique and soft-tissue coverage in Dupuytren’s disease. Hand Clin 1999;15:127.

McFarlane RM. Severe contractures of the proximal interphalangeal joint in Dupuytren’s disease: combined fasciectomy with capsuloligamentous release versus fasciectomy alone – discussion. Plast Reconstr Surg 1996;97:567.

McKenna SP, Wilburn JN, Crawford SR, Perry-Hinsley D, Bayat A. Dupuytren’s disease and quality of life. Wound Repair Regenerat 2012;20:A31.

Raina S, Jaryal A. Dupuytren’s contracture. Indian J Med Res 2012;136:502.

Raven RB, III, Kushner H, Nguyen D, Naam N, Curtin C. Analysis of efficacy and safety of treatment with collagenase clostridium histolyticum among subgroups of patients with dupuytren contracture. Ann Plast Surg 2013:73:286–90.

Salhi S, Cardin-Langlois E, Luc M. Percutaneous fasciotomy for the treatment of Dupuytren’s disease - a systematic review. Hand 2011;6:349–55.

Stanbury SJ, Hammert WC. Dupuytren Contracture. J Hand Surg Am 2011;36:2038–40.

Townley WA, Baker R, Sheppard N, Grobbelaar AO. Dupuytren’s contracture unfolded. Br Med J 2006;332:397–400.

Tropet Y, Deck D, Vichard P. Dupuytren’s contracture of the little finger. [French]. Ann Chir Main Memb Super 1994;13:101–6.

Varian JPW, Hueston JT. Occurrence of Dupuytren’s disease beneath a full thickness skin graft: a semantic reappraisal. Ann Chir Main Memb Super 1990;9:376–8.

Wegman T. Treatment of Dupuytren’s disease with collagenase. Official Gazette of the United States Patent and Trademark Office Patents 1996;1193:3493.

Wilbrand S, Ekbom A, Gerdin B. The sex ratio and rate of reoperation for Dupuytren’s contracture in men and women. J Hand Surg 1999;24:456–9.

Wilbrand S, Ekbom A, Gerdin B. Dupuytren’s contracture and sarcoma. J Hand Surg 2002;27:50–2.

No relevant outcomes (n = 6)

Badalamente MA, Hurst LC, Ward A, Parker T, Liu J, Tatka J. Histology of cords from Dupuytren’s patients: a comparison after collagenase injection/fasciectomy, primary fasciectomy and repeat fasciectomy: not a clinical study. J Hand Surg Am 2012;37:08.

De Salas-Cansado M, Ruiz-Antoran MB, Ramirez E, Abad SF. Healthcare resource utilization and associated costs with surgical treatment of Dupuytren disease in Spain. Basic Clin Pharmacol Toxicol 2001;109(Suppl. 3):33.

De Salas-Cansado M, Ruiz-Antoran MB, Ramirez E, Abad SF. Healthcare resources utilization and associated costs with surgical treatment of Dupuytren’s disease in Spain. Basic Clin Pharmacol Toxicol 2011;109:37.

Messore L, De DC, Battaglini G, Pizza G, Novelli A, Franculli F. Dupuytren disease. Remarks in surgical technique. [Italian]. Minerva Orthop Traumatol 1998;49:165–8.

Skodny P, Mackowiak JI, Peimer C. Efficacy and effectiveness of collagenase clostridium histolyticum for Dupuytren’s contracture. Value Health 2011;14:A302.

Starkweather KD, Lattuga S, Hurst LC, Badalamente MA, Guilak F, Sampson SP, et al. Collagenase in the treatment of Dupuytren’s disease: an in vitro study. J Hand Surg Am 1996;21:490–5.

No translation (n = 2)

Eisenschenk A. Collagenase therapy for Dupuytren. Unfallchirurg 2011;114:965.

Anonymous. Dupuytren needle fasciotomy for primary treatment of Dupuytren’s contracture. Unfallchirurg 2012;115:765.

Pre-1990 (n = 1)

Norotte G, Apoil A, Travers V. A ten years follow-up of the results of surgery for Dupuytren’s disease. A study of fifty-eight cases. Ann Chir Main 1988;7:277–81.

Unable to obtain (n = 1)

Donato De FA, Pardini Junior AG, Tufi Neder FA. Dupuytren’s contracture: treatment using the open palm technique. [Portuguese]. J Braz Orthoped 1997;32:301–4.