Total hip replacement and surface replacement for the treatment of pain and disability resulting from end-stage arthritis of the hip (review of technology appraisal guidance 2 and 44): systematic review and economic evaluation

Osteoarthritis of the hip

The hip is a weight-bearing ball and socket joint that is commonly affected by OA. OA in the hip manifests itself as loss of articular cartilage, inflammation of synovial tissue and hypertrophy of the associated bone (e.g. osteophytes, bone sclerosis). The loss of cartilage tissue and new bone tissue growth suggests OA may result from disordered repair of cartilage damaged by mechanical and biochemical changes within the joint. 3

When the repair process is unable to keep up with the rate of tissue damage, the consequence is symptomatic OA characterised by pain, stiffness and progressive disability. 3

Osteoarthritis of the hip may be classified as primary or secondary. Secondary hip OA can be caused by most intra-articular diseases, including osteonecrosis, trauma, septic arthritis, Paget’s disease, hip dysplasia, Perthes’ disease and slipped upper femoral epiphysis. Primary hip OA is presumed when no other specific cause has been identified. 3

Rheumatoid arthritis of the hip

Rheumatoid arthritis (RA) is an autoimmune disease that commonly affects the synovial lining of peripheral joints, including those of the hand, foot and hip. RA is a multisystem disorder with implications for almost every region of the body, including the heart, lungs and eyes. 4 Multiple episodes of synovial inflammation lead to reduced articular cartilage (e.g. causing secondary OA), joint destruction and progressive disability. It has also been associated with reduced quality of life and premature mortality. 5–7

Rheumatoid arthritis manifests itself by gradual accumulation of structural changes within the joint, which can (particularly in late-stage disease) be detected by radiography or other imaging techniques. 5 In 2010, a joint working group of the American College of Rheumatology and the European League Against Rheumatism5 developed new criteria for identifying patients with early RA, which place more emphasis on characteristics associated with a high risk of later progression to severe and erosive disease.

Epidemiology of osteoarthritis and rheumatoid arthritis

Osteoarthritis is one of the most commonly encountered musculoskeletal diseases. There are an estimated 2.8 million patients with OA in the UK, based on symptomatic diagnosis in patients aged > 45 years. 8 A further 8.5 million people are estimated to be affected by joint pain that can be attributed to OA. 3

Current projections estimate that 10% of the world’s population aged ≥ 60 years will present with symptoms caused by OA. 9 The prevalence and incidence of OA, including hip OA, increase with age and are higher in women than in men after 50 years of age. 10,11 For example, the incidence rates of hip OA in men and women aged 70–79 years are estimated to be 430 and 600 per 100,000 person-years, respectively. 12

Estimates of age-standardised incidence rates of hip OA among women and men in Europe are about 53.3 and 38.1 per 100,000, respectively. 13 The prevalence of hip OA among Caucasians is demonstrably higher (range 3–6%) than in Asian, black and East Indian populations (≤ 1%). 14 In light of a longer life expectancy, an ageing population and increasing rates of obesity observed in developed countries, it is expected that both the incidence and the prevalence of OA will rise in future. 1,15,16

It is difficult to estimate the prevalence and incidence rates of OA accurately because of variable diagnostic criteria (e.g. radiographic, symptomatic or self-reported features). 10,17,18 For example, some patients with radiographic evidence of joint damage indicative of OA may not experience pain or disability whereas some patients with clinical OA may not demonstrate radiographic changes. These discrepancies make it challenging to determine the presence or absence of OA accurately. 10 In general, the prevalence of symptomatic or self-reported OA is higher than that of radiographic OA. 3

The prevalence of RA is estimated at 400,000 cases in the UK. Estimates of annual incidence suggest that 10,000–20,000 people develop RA in the UK each year. Although the disease may develop in patients at any age, onset classically occurs between the ages of 40 and 60 years. The incidence of RA is approximately two to three times greater in women than in men4 and approximately 10–40% of cases manifest within the hip. 19

Risk factors for osteoarthritis

Evidence suggests that contributing factors to OA can be classified broadly as:

1. biomechanical (e.g. joint injury, reduced muscle strength)

2. constitutional [e.g. advanced age (≥ 65 years), female sex, obesity and high bone density]

3. genetic (high heritability estimates for OA).

Biomechanical factors are probably the most important cause and may explain both the relationship between OA and obesity as well as the tendency for OA to affect weight-bearing joints, for example the hips and knees. 2 Malalignment, instability and altered joint loading correlate with OA progression in both clinical and animal studies. 20,21 In the hip, femoroacetabular impingement are related to OA onset; ‘cam type’ is a bump on the surface of the femoral head typically affecting younger athletic men and ‘pincer type’ impingements describe an overdeep acetabululm, which restricts the movement of the femoral head – this typically affects middle-aged women. The prevalence of any type of congenital or acquired hip malformation is 4.3% in men and 3.6% in women. Similarly, epidemiological studies have demonstrated associations between certain occupational factors (e.g. long-distance running, farming, heavy physical work load) and hip OA. 22,23

However, biomechanical factors alone do not explain the onset of OA in non-weight-bearing joints (e.g. the carpometacarpal joints) and metabolic factors may also play a role. 2,24

Symptoms and diagnosis

Symptoms of hip OA include pain, stiffness and loss of function, that is, limited daily activities such as walking, climbing the stairs and performing household tasks. 1,11,19,25 The diagnosis of primary hip OA is usually based on history and clinical examination with particular assessment of joint pain, deformity and reduced range of movement. Physical examination can also exclude pain resulting from other causes, for example bursitis, tendonitis and muscle spasm. Plain radiographs of the hip are used to identify and stage OA.

Advanced imaging techniques such as magnetic resonance imaging (MRI) and computerised tomography can identify causes of secondary hip OA (e.g. stress fractures, osteonecrosis, Paget’s disease, inflammatory arthropathies) as well as evaluating and monitoring the extent of hip damage. 1,18

Natural history of osteoarthritis

The natural history of OA varies between affected joints but little is known about the natural history of the symptomatic disease. The prognosis of hip OA has been shown to be the least favourable and is the most frequent reason for surgical intervention after 1–5 years of progression. 3 The national clinical guideline (CG) for OA3 states that hip OA has the worse outcome of all the OA sites discussed in the CG (hip, knee, hand). Occasionally, OA hips can improve without surgical intervention as measured by symptoms and radiographic change. 3 Comorbidity (e.g. diabetes, obesity, cardiovascular disease) may additionally influence the prognosis of OA, as does older age. 3

Significance for patients in terms of ill health (burden of disease)

Osteoarthritis has a significant impact on an individual patient, resulting in pain, stiffness, limited mobility and reduced function. A UK-based survey assessed the impact of OA on daily living for 1762 people. 26 The majority of the sample consisted of people aged ≥ 50 years, of whom 75% were female. In total, 81% of respondents were found to have experienced constant pain and/or were limited in their ability to perform everyday tasks. Many respondents had visited a general practitioner three or four times before a diagnosis of OA, which was made on average 18 months after the onset of symptoms. Approximately 72% of respondents had comorbid conditions such as heart disease, diabetes and hypertension.

Significance for the NHS

The economic impact of arthritis consists of direct costs to health-care services and indirect costs because of lost productivity and early mortality. The impact of OA on health services and the UK economy has been substantial. The cost of treating OA has been estimated to be approximately £640 per person per year. 19 A report has suggested that, if one-tenth of the 15.2 people per 1000 who experience hip pain severe enough for surgery received medical and/or physical therapy, the cost to the NHS in England and Wales would be of the order of £48M per year in 2002. 19 The costs of both surgical and non-surgical interventions are reviewed in detail later in this chapter.

Because of the ageing of the population, OA is projected to become the fourth leading cause of disability worldwide by 2020. 3 In the present economic climate of tightening health-care spending, the implications of increasing demand for the treatment of arthritis of the hip have led to intense discussions about the cost-effectiveness of new technologies and treatment options.

Measurement of disease

More than 20 tools have been developed and validated for the assessment and monitoring of patient outcomes specific to hip arthritis. 27 One commonly used disease-specific tool is the Western Ontario and McMaster University Osteoarthritis Index (WOMAC). 28 This is a 24-item questionnaire that covers three domains of pain, stiffness and physical function, with a total score ranging from 0 (worst outcome) to 100 (best outcome). Other validated tools designed to measure outcomes specific to hip function and symptoms (e.g. disability, pain, range of motion, limitations in daily living and other activities) have also been used. 27,29

In the UK the most commonly used tools are the Oxford Hip Score (OHS)30 and the Harris Hip Score (HHS). 31

Management of disease

Treatment and management of arthritis in the UK can be categorised as non-surgical and surgical as detailed below. Patients in the early stages of OA begin treatment with non-surgical options; when non-surgical management has failed, patients are considered for intervention with surgical treatment.

Non-surgical management:

• self-management and patient education

• non-pharmacological (acupuncture, exercise, physical therapy, manual therapy, weight reduction)

• pharmacological [simple analgesics, non-steroidal anti-inflammatory drugs (NSAIDs), topical treatments, intra-articular steroid injections].

Surgical management:

• surgery [e.g. THR or resurfacing arthroplasty (RS), arthrodesis, arthroscopy, osteotomy].

Current service cost

Arthritis has a significant negative impact on the UK economy with an estimated total cost of 1% of gross national product. 36 It is the most common group of conditions for which people receive Disability Living Allowance in England. The benefits provided outweigh those provided for people diagnosed with heart disease, stroke, chest disease and cancer combined. 36 A reported £43M is spent annually on community services and £215M on social services for OA. 36 In 2002 an estimated 36 million workdays were lost because of OA, resulting in £3.2B of lost productivity. 36 Data for the numbers of people who have their symptoms managed by non-surgical interventions (such as pain relief, exercise, physical therapy and manual therapy) within England and Wales are difficult to ascertain.

Chen et al. 8 estimated the cost of topical and oral NSAIDs using prescribing data from 2005/6. They reported that an estimated 167,000 people with a diagnosis of OA were found to have been prescribed topical NSAIDs and 1.4 million patients were prescribed oral NSAIDs. The annual costs were £8.5M and £25M, respectively. 8 Adjusting for inflation they found that this would equate to £19.2M and £25.65M, respectively in 2010. Most health economic analyses have reported that surgery for the treatment of arthritis is a cost-effective intervention and maximises cost per quality-adjusted life-year (QALY) gained. 37

An earlier Health Technology Assessment (HTA) report (reference number 01/21/01)19 found that the annual cost to the NHS of elective hip replacement surgery for the treatment of OA was £140M and that each trust spent, on average, £257,000 on the purchase of hip prostheses in 1998/9. This study was conducted in 2002. 19 It reported that the cost to the NHS and social services of non-surgical treatment for an individual was approximately £640 per person per year. During the year 2000, £405M was spent on 44,000 hip and 35,000 knee replacements. 36 Since then the costs have increased substantially, as the estimated cost to the NHS of THR surgery alone in 2011 was reported to be £426M. 36

The cost of one surgical treatment in 2002 was £3891, averaged across all NHS trusts in 1999/2000, with the cost for 50% of trusts falling within the range £3404–4434.23. 19 According to the 8th Annual Report of the National Joint Registry for England and Wales (NJR),36 the cost of hip replacement surgery varies considerably from trust to trust in the UK, with no set national price for implants. The cost depends considerably on length of hospital stay. For example, the tariff reimbursement paid to a trust in one study in 2005/6 for a primary THR was £6000 whereas, in 2010, the national tariff was set at £5552 for an uncomplicated THR. 36

When hip replacement surgery fails, revision surgery to replace part or all of the prosthetic hip joint may be required. The number of revision procedures has increased in recent years, with 3012 carried out in 2003/4, rising to 6581 by 2008/9. 36 This accounted for approximately 9.4% of all elective hip replacement procedures performed in England and Wales. 36 Revision surgery is also a key element of the current service expenditure, with unit costs of revision generally higher than those for primary surgery. Briggs et al. 38 reported a mean cost for a standard hip revision procedure in 2000/1 as £5294 (£6385 in 2008 prices) compared with £3889 (£4690 in 2008 prices) for a primary procedure. The 2002 HTA report19 stated that, in 1989/90, one in seven of all procedures (5000 out of a total of 35,000) was a revision of a hip replacement. In 1999/2000 a crude estimate of 6700 revisions was reported. 19

Randomised controlled trials (RCTs) have compared revision rates across prosthesis types but with insufficient sample sizes or durations of follow-up to produce conclusive results. 39 The largest observational study found that 7-year revision rates were lower for cemented (3.0%) than for hybrid (3.8%) or cementless (4.6%) prostheses. 36 Edlin et al. 40 reported that a total of 97% of UK hip replacements are still working (unrevised) at 5 years.

Variation in services and uncertainty about best practice

Outcomes for hip replacement surgery vary by geographical location, surgeon and hospital. The Global Orthopaedic Registry has shown that patient selection criteria vary between practitioners, surgeons and referring doctors and between countries. 41 Nationally, there are reported inconsistencies in the treatment, procedure and prostheses that are offered to patients in the NHS. 42

In 1998 more than 60 hip prostheses manufactured by 19 companies were available commercially in the UK, with total NHS expenditure of approximately £53M. 43 By 2008 this had risen to 124 brands of acetabular cups and 137 brands of femoral stems at a cost of £67M. 36 This represents a substantial increase in the variety of available prostheses in recent years. Implants are often grouped into cemented, cementless and hybrid prostheses. 44 The reported increasing use of cementless components in the UK has contributed to a doubling of prosthesis costs between 1996 and 2006. 44

There is variation in the rate of primary hip replacement expenditure in England per 1000 population weighted by age, sex and need. For example, hip RS accounts for 6% of the approximate 70,000 hip arthroplasty operations conducted in England and Wales every year, although the equivalent figure among men aged < 55 years is 33%. 40

Spend also varies significantly between regions in the UK, with the lowest reported in Tower Hamlets (£560) and the highest in Devon (£8140). 42 When examining data by local authority, the difference in the rate of provision of hip replacements per 1000 people in need was almost 14-fold. 42 National European Quality of Life-5 Dimensions (EQ-5D) data after hip replacement for England and Wales show that variation between the best and worst trusts is large (31–49%) and cost-effectiveness varies considerably between hospitals. 45

Summary of National Institute for Health and Care Excellence guidance on the selection of prostheses for primary total hip replacement

The 2000 technology appraisal (TA)246 stated that the ‘benchmark’ for the selection of prostheses for THR should be a revision rate of ≤ 10% at 10 years with evidence relating to data from adequately sized, well-conducted observational studies or RCTs. NICE recommended that various patient factors, including age and underlying pathology, should be taken into account when choosing prostheses, for example ease of revision (of particular importance for younger patients).

Specific recommendations on the selection of hip prostheses for primary THR were considered difficult to construct because the evidence base was generally poor and difficult to interpret. However, the available evidence supported the use of a range of cemented prostheses for primary THR. This was further supported by the evidence on immediate and long-term postoperative pain.

There are currently no cost-effectiveness data based on revision rates after ≥ 10 years of follow-up to support the use of the generally more costly cementless and hybrid hip prostheses. Some evidence suggested that these types of prostheses might lead to less bone loss, meaning that they were potentially easier to revise than cemented prostheses. However, no reliable evidence was available to support the proposition that the potential ease of revision of a hip prosthesis would outweigh its poorer revision rate.

Summary of National Institute for Health and Care Excellence guidance on the use of metal-on-metal hip resurfacing arthroplasty

In the June 2002 NICE guidance TA44,25 metal-on-metal hip RS was recommended as one option for people with advanced hip disease who would otherwise receive, and are likely to outlive, a conventional primary hip replacement. It did note, however, that the current evidence was principally in individuals aged < 65 years and that surgeons should bear this in mind. Furthermore, the guidance stated that all patients receiving this arthroplasty should be made aware of the relative paucity of evidence for medium- to long-term safety and reliability and the likely outcome of revision surgery compared with that for conventional THR.

However, in June 2012 advice about follow-up of patients receiving a metal-on-metal articulation changed. The Medicines and Healthcare products Regulatory Agency (MHRA) issued a medical device alert47 stating that a small number of patients implanted with these hips might be at risk of developing progressive soft tissue reactions to the wear debris associated with metal-on-metal articulations; this updated the original advice of April 2010. These reactions could also adversely affect the results of later revision surgery. However, it also stated that its evidence pointed to the fact that early revision of such poorly performing metal-on-metal articulations should give a better revision outcome. Therefore, the agency advised that clinicians should perform appropriate follow-up, depending on which group a patient’s hip surgery fitted into, as well as whether the patient was symptomatic or asymptomatic. Follow-up, if indicated, should consist of both imaging (MRI or ultrasound) and blood metal ion tests [ion level greater than seven parts per billion indicates the potential for soft tissue reaction]. Revision should be considered if imaging is abnormal and/or blood metal ion levels are rising.

Summary of Medicines and Healthcare products Regulatory Agency alert advice

Metal-on-metal hip RS implants:

• symptomatic: follow-up annually for life of implant

• asymptomatic: follow-up according to local protocols – no need for investigations unless cause for concern about cohort or patients who become symptomatic.

Metal-on-metal THRs with a head diameter < 36 mm:

• symptomatic: follow-up annually for life of implant

• asymptomatic: follow-up according to local protocols – no need for investigations unless cause for concern about implant.

Metal-on-metal THRs with a head diameter ≥ 36 mm:

• annual follow-up for life of implant whether symptomatic or not.

DePuy ASR^™ hip replacements (all types) (DePuy, West Chester, PA, USA):

• annual follow-up for life of implant whether symptomatic or not.

National Institute for Health and Care Excellence guidance on the care and management of osteoarthritis in adults

The most recent NICE guidance on OA, issued in February 2008,3 stresses the importance of a holistic assessment of the patient, including his or her function, quality of life, occupation, mood, relationships and leisure activities. After this assessment, the clinician is advised to formulate and agree a management plan with the patient, which should include ‘core treatments’ such as education, muscle strengthening and aerobic exercise, and weight-loss programmes for the overweight or obese. It should also include other self-management and ‘conservative’ strategies such as application of heat/cold packs or transcutaneous electrical nerve stimulation to the site of pain, manipulation and stretching (particularly for hip OA) and assessment for bracing/joint supports/insoles/walking sticks.

Adjuncts to the above ‘core’ treatment could include pharmacological treatments, in particular paracetamol (regular dosing may be required) and topical NSAIDs or topical capsaicin (although topical treatments are less useful for hips). If these are found to be insufficient for relieving pain, practitioners are advised to consider adding opioid analgesics or oral NSAIDs. Intra-articular corticosteroid injections are recommended for moderate to severe pain. Clinicians are advised to consider a referral for joint surgery if the patient has already been offered the ‘core’ treatments and is still experiencing joint symptoms that have a substantial impact on quality of life.

The Orthopaedic Data Evaluation Panel

The Orthopaedic Data Evaluation Panel (ODEP) was established to provide an independent assessment of clinical evidence, submitted by suppliers, on the compliance of their implants for THR and hip RS with NICE benchmarks for safety and effectiveness. ODEP produced detailed criteria for this assessment and in 2010 there was an ongoing review of this guidance by all stakeholders. 36 ODEP does have to rely on the honesty of the submitting companies and therefore provides no warranty that the data in its database are accurate, complete or current.

For 10-year benchmark products (those recommended to last for 10 years), ODEP places them in one of four categories according to whether there is evidence that a product meets NICE guidelines:

• level A – strong evidence that product meets NICE guidelines

• level B – reasonable evidence that product meets NICE guidelines

• level C – weak evidence that product meets NICE guidelines

• unacceptable – unacceptable evidence that product meets NICE guidelines.

For products that fail to meet NICE’s 10-year benchmark, ODEP looks at evidence at 3, 5 and 7 years. Again, these products are split according to whether there exists acceptable, weak or unacceptable evidence for the product meeting NICE guidelines.

As of March 2011, ODEP ratings had been given to 38% of available brands of femoral stems and 41% of available brands of acetabular cups used in primary procedures. However, 42% of available brands of acetabular cups and 47% of available brands of femoral stems being used in England had not yet submitted data to ODEP. Clearly, for surgeons to make the most informed choices, it is important that all manufacturers submit their product data to ODEP using the pro forma and associated guidelines.

Example patient care pathway for hip arthroplasty

Figure 1 presents a typical care pathway for patients treated for arthritis in the NHS. In general, patients would be treated in primary care services and undergo various non-surgical management options. Once non-surgical management is said to have failed, the patient is classified as having end-stage arthritis and is recommended for surgery in secondary care.

FIGURE 1.

Example pathway for patient with arthritis in primary care.

Figure 2 presents the two surgical options THR and hip RS. The care pathways are similar in terms of pre- and postoperative care and follow-up.

FIGURE 2.

Example hip replacement care pathway in secondary care.

Identification of different types of total hip replacement

The different types of THR can be categorised into the following subgroups:

• hip replacement with different fixation methods for implant components (cemented, cementless, hybrid or reverse hybrid prostheses)

• hip replacement with implant components (i.e. femoral stem, femoral head, acetabular cup) made from different materials (metal, ceramic, polyethylene)

• hip replacement with differing femoral head sizes.

Hip replacement with different fixation methods

Hip replacement prostheses can be categorised by their fixation method (Figure 3) as (a) cemented, (b) cementless, (c) reverse hybrid with a cemented cup and cementless stem or (d) hybrid with a cemented stem and cementless cup. Cemented prostheses are held in place with bone cement and generally consist of three components: a femoral stem, a femoral head (modular) and an acetabular cup. These components are permanently attached to the pelvis and the femur. According to the NJR, the percentage of cemented procedures did not change between 2009 and 2010. The number of cemented procedures had been in decline since 2005. In 2004 the figure was at 77%, and by 2010 this had reduced to 50%. 36

FIGURE 3.

Overview of four different fixation options for the femoral stem and acetabular cup in THR. (a) Cemented THR; (b) cementless THR; (c) reverse hybrid THR – cementless stem with a cemented cup; and (d) hybrid THR – cemented stem with a cementless cup.

Cementless prostheses rely on initial press-fit fixation followed by natural bone growth. They typically consist of four components: a femoral stem, a femoral head, an acetabular cup shell and an acetabular liner. The theoretical benefit of the cementless fixation is the possibility of bone–implant interface (human : technology) remodelling. In England and Wales there has been a 4% increase in cementless procedures in recent years. 36

The cementless prostheses include implant components coated in a porous material (hydroxyapatite) that is compatible with bone growth and which helps to secure the liner in place. Hydroxyapatite is a mineral form of calcium apatite. 50 Hydroxyapatite is also commonly used as a filler to replace amputated bone in addition to a coating to promote bone ingrowth into prosthetic implants.

A hybrid hip replacement consists of a cemented femoral stem and a cementless acetabular cup, whereas the reverse hybrid uses a cementless femoral stem and a cemented acetabular cup. In 2010, 14% of these types of procedure were reverse hybrid (cementless stem, cemented acetabulum) and 86% were standard hybrid (cemented stem, cementless acetabulum). 36

Hip replacement with components made from different materials

The combinations of prosthetic components that are available are listed in Table 1. The different materials used for the implant components (i.e. femoral stem, femoral head, acetabular cup) produce various articulating surfaces or bearing surfaces.

The NJR report for 201136 provided the percentage use of fixation type during 2010 and 2011 (Table 2). The cemented fixation type was the most popular fixation method and the polyethylene-on-metal articulation combination was used the most (86.1%) of all the cemented bearing surfaces. The cementless fixation type was the second most common fixation method and the polyethylene-on-metal articulation combination was most popular (35.6%).

Another way of characterising the variation of combination of articulation surface and fixation method is by frequency of use, as reported in the NJR. The most common combinations are listed in Table 3 along with the associated acronyms that have been used in the remainder of this report.

Polyethylene-on-metal (cup material-on-femoral head material)

A metal ball with polyethylene cup (or polyethylene liner inside a metal cup) (Figure 4) is the most common type of articulation combination (both cemented and cementless) and is one of the cheapest. The Charnley low-friction arthroplasty was the first widely accepted polyethylene-on-metal prosthesis to be used. It has a high reported implant survivorship at > 20 years’ follow-up (> 80%) and at 35 years’ follow-up (78%). 41 It also provides the baseline against which new prosthetic designs are compared. In England and Wales this was the most common articulation type used during 2010 and 2011 (see Table 2). Clinical advice suggested that, if a metal cup is used with a polyethylene liner, a cementless cup fixation is most commonly used in England, and the cementing of the metal cup is increasingly rare. Highly cross-linked polyethylene is being used by some surgeons in place of standard polyethylene in THRs because of its lower reported wear rates. 51,52

FIGURE 4.

Cemented metal stem, metal femoral head and polyethylene acetabular cup.

Polyethylene-on-ceramic

The polyethylene-on-ceramic option combines a polyethylene cup with a hard ceramic femoral head (Figure 5). This articulation type is reported to have a lower wear rate than the polyethylene-on-metal combination and is cheaper than the ceramic-on-ceramic option. It is used more often with a cementless fixation (14.2%) than with a cemented fixation (8.4%) (see Table 2). The ceramic head is harder than metal and hence reportedly withstands more wear. In the past ceramics were brittle and cracked, leading to failure of the implant, but advances in technology have limited this problem in recent years.

FIGURE 5.

Cemented metal stem, ceramic femoral head and polyethylene acetabular cup.

Ceramic-on-ceramic

The ceramic-on-ceramic articulation (Figure 6) provides the hardest bearing surface combination and is generally the most expensive combination available. 40 This combination has a lower reported wear rate than other options available to patients in England and Wales. The ceramic-on-ceramic articulation is mostly used without cement, as shown in Table 2 (25.6% cementless vs. 1.8% cemented). Clinical advice suggests that the cementless ceramic cup is the most common type of ceramic-on-ceramic articulation in England; cementing the ceramic cup is increasingly rare, as demonstrated in the NJR data. 36

FIGURE 6.

Cemented metal stem, ceramic femoral head and ceramic acetabular cup.

Metal-on-metal

Metal-on-metal articulations (Figure 7) provide a hard bearing surface; however because of their reportedly high revision rate the MHRA has made recommendations for following up patients implanted with such devices. 47

FIGURE 7.

Cemented metal stem, metal femoral head and metal acetabular cup.

The MHRA recommendations apply to four groups of metal-on-metal replacements:

1. metal-on-metal hip RS implants

2. metal-on-metal THRs with a head diameter < 36 mm

3. metal-on-metal THRs with a head diameter ≥ 36 mm

4. DePuy ASR hip replacements comprising:

• ASR acetabular cups for hip RS or THR

• ASR surface replacement heads for hip RS

• ASR XL femoral heads for THR.

Revision is necessary when prostheses fail, more common in younger patients, usually for loosening secondary to wear or dislocation. Interestingly, metal-on-metal bearing surfaces were actually designed by surgeons to reduce the proportion of replacements that require revision. They had been extensively assessed in simulator tests and were noted to be highly resistant to wear, even when used in very large head sizes. 53

Head size is important because in simulator tests larger head sizes give lower wear because of the boundary lubrication regime becoming more favourable. 54 Therefore, implantation of large diameter metal-on-metal bearing surfaces on stemmed prostheses became popular on the basis of such evidence, which suggested that they should result in less wear and thus lower failure rates. They seemed to be particularly appropriate for younger, more active patients.

However, several issues have arisen with the practical use of these metal-on-metal prostheses. It soon emerged that one brand of metal-on-metal prosthesis, the DePuy ASR, actually seemed to fail early. 55 Data received from the company56 showed that 5 years after surgery 12% of patients who received the ASR RS and 13% of patients who received the ASR THR required revision surgery.

This prompted recent analysis of NJR data on 402,051 hip replacements to assess whether metal-on-metal bearing surfaces lead to increased implant survival compared with other bearing surfaces in stemmed THR. 16 These authors additionally challenged the previous evidence that larger head sizes result in improved implant survival.

The results revealed that, in THR, metal-on-metal articulations failed at higher rates than other bearings. For example, 5-year revision rates in younger women were 6.1% [95% confidence interval (CI) 5.2% to 7.2%] for 46-mm metal-on-metal articulations compared with 1.6% (95% CI 1.3% to 2.1%) for 28-mm polyethylene-on-metal articulations. This effect was found even though the ASR data had been removed before analysis (the DePuy ASR articulations had already been removed from the market). Thus, it is a problem with all metal-on-metal prostheses, not an implant-specific characteristic. In addition, their failure was found to be related to head size, with larger heads failing earlier than smaller versions (this effect was the opposite to that found for ceramic-on-ceramic articulations). The authors suggested a number of potential reasons for the finding that larger metal heads fail earlier, such as a failure to achieve optimum lubrication or trunnion (post that inserts into head) wear55 resulting in metal debris leading to local soft tissue reactions57 or early loosening because of increased transmitted torque from the larger head. The authors of the paper therefore suggested that metal-on-metal replacements not be performed because of poor implant survival and that patients undergo at least an annual review with both clinical and radiological examination, in line with the MHRA recommendations. 47

Furthermore, there are the potential dangers of exposure to metals such as chromium and cobalt. Metal alloys used in metal-on-metal bearings degrade through wear, from corrosion or because of a combination of the two. 58 Consequently, they produce a vast number of nanometre- to submicrometre-sized metal particles that cumulatively present a large surface area for corrosion. 59 This is also relevant to the polyethylene-on-metal bearings, which also produce such particles through wear. The consequences of local and systemic exposure to the wear particles and the accompanying biologically active corrosion products have been extensively researched. 60 It is well known that metal debris can induce adverse local soft tissue reactions41 including the release of inflammatory cytokines from macrophages, histiocytosis, fibrosis and necrosis. 61 Local results include aseptic loosening because of osteolysis induced by some immunological reaction involving hypersensitivity62 and local pseudotumours (soft tissue masses relating to the joint) that are locally destructive and require revision surgery in the majority of patients. 63

Furthermore, it seems that metals can disseminate through the body and cause direct damage to end organs such as the kidneys, lungs and brain. 64,65 There is also evidence of genotoxicity and evidence that these metals can signal across biological barriers at concentrations produced after THR. 66 The genotoxic effects of the metal ions are thought to be mediated either by direct action, causing DNA breaks through attacks on free radicals, or through an indirect effect by inhibiting the repair of DNA. 67 There have been concerns that this genotoxicity could cause a long-term increased risk of malignancy, particularly important for the younger, more active patients in whom life expectancy after implantation is long. However, recent studies have failed to find this increase68 and some have actually found a decrease in the numbers of certain malignancies in metal-on-metal articulation patients. 69

The US Food and Drug Administration (FDA),70 the UK MHRA47 and the British Orthopaedic Association71 have released statements of concern about metal-on-metal articulations. The MHRA recommendation states that patients with metal-on-metal bearings and a painful hip joint should have yearly measurements of whole blood metal ion concentrations and radiographic assessment to exclude adverse local tissue reactions as the source of pain. 47 These yearly assessments should continue for the lifetime of the hip replacement. The use of metal-on-metal bearing surfaces has consequently declined in England and Wales. In 2010/11 only 7.9% of all procedures used a metal-on-metal implant (see Table 2). However, data suggest that they are still being used extensively in other countries. For example, in the USA, 35% of articulations were metal-on-metal in 2009. 72

Hip replacement with differing femoral head sizes

Research has suggested that differing femoral head sizes leads to variation in the rate of revision. Smith et al. 16 reported that the use of larger head sizes (> 36 mm in diameter) improves stability and range of motion compared with the smaller head diameters that are used with other bearing surfaces. Use of large diameter femoral heads increases the distance that the head must travel before dislocation, without decreasing hip range of motion, thus increasing stability. 41

Implant instability and dislocation

Instability and recurrent dislocation are the most common reasons for THR failure and the second most common cause of failure of revision THR. The prevalence of dislocation ranges from 0.3% to 10% for primary THR and is 28% for revision THR. 75–77

The most common reasons for instability are component malpositioning and abductor (muscle) deficiency such as a loss of abduction power, which can lead to a severe limp. Cup malpositioning can lead to increased wear of particular sections of the prosthesis, for example both 45-degree inclination (tilting) and 20-degree anteversion (forward tilting) have been associated with THR failure. 78,79 However, age, previous fracture, surgical volume, surgical approach, component sizing and polyethylene wear are also contributory factors to revision because of instability and dislocation. 80–83

Recurrent late dislocation remains a major source of THR failure. There are various treatment options for patients who have recurrent dislocations. These include revision surgery using constrained polyethylene liners (which offers increased stability but at the cost of smaller range of motion), larger diameter femoral heads and dual mobility devices.

Aseptic loosening and osteolysis

Aseptic loosening is a common cause of failure of THR. It arises because of osteoclast-mediated bone reabsorption at the bone–implant interface, which can lead to loosening, implant migration, implant failure and periprosthetic fracture. 84 Osteolysis is one of the most common complications after THR, which may lead to implant failure. It is initiated as a result of an inflammatory process against polyethylene particulate debris. Component malpositioning is a major cause of severe wear and osteolysis, but they are also affected by activity level and material and component design. 85

Aseptic loosening and osteolysis are diagnosed clinically by patient reports of pain. They are treated with replacement of loose components and correction of component malalignment. Outcomes after revision surgery are generally good, with reported mechanical failure rates < 5% at follow-up. 86

Periprosthetic fracture

Periprosthetic fracture is a major complication after THR and is associated with increased morbidity and mortality. Risk factors for periprosthetic fracture include previous revision surgery, component malalignment, age, osteoporosis, previous fracture and minor trauma. 87,88

Treatment for most periprosthetic fractures is usually surgical. Treatment for most periprosthetic fractures is usually surgical and the options depend on the fracture pattern. It can include open reduction and internal fixation with or without cortical strut allografts, longer femoral stems or increases in the setting of acetabular fractures, or tumour prostheses. 89,90

Infection

Infection of a THR prosthesis is associated with greatly increased morbidity, mortality and use of health-care resources. The infections can by treated with antibiotics; however, deep infections are rarely cured by antibiotics alone and may require revision surgery. As more THRs are performed, the absolute number of deep infections is likely to increase although, because of comprehensive infection control techniques, rates are relatively low. Risk factors for infection include age, obesity, comorbidities and American Society of Anesthesiologists (ASA) score. Longer operative times and reoperation within 90 days have been implicated as risks for infection. 91,92

Clinical follow-up

Implants should be assessed every year for signs of loosening, migration/measure of prosthesis movement (e.g. femoral head penetration rate) and failure. Although no studies have examined the benefits of specific follow-up frequencies, NICE recommends continued periodic follow-up.

Follow-up using radiostereometric analysis allows for precise quantification of any implant movement of the prosthesis; however, visual inspection of the radiograph by the surgeon is commonly used in clinical follow-up. 93 Evidence suggests that early detection of lesions (e.g. aseptic lymphocyte-dominated vasculitis) is more cost-effective than waiting until patients report pain and loss of function and an assessment is conducted. 94

Disability, function, pain, limitations in daily activities, overall satisfaction and health-related quality of life should be routinely measured and documented at follow-up using validated instruments [e.g. Short Form questionnaire-12 items/Short Form questionnaire-36 items (SF-12/SF-36), EQ-5D]. 27

Search strategies

Final searches were undertaken in November and December 2012 (see Appendix 1) and were date limited from 2002 (the date of the most recent NICE guidance in this area25). Searches of the clinical effectiveness literature were restricted to RCTs and systematic reviews; additional searches were undertaken to capture literature relating to costs, resource use, utilities, cost-effectiveness, cost-effectiveness models and registries to inform the survival and cost-effectiveness analysis.

The following main sources were searched to identify relevant published and unpublished studies and studies in progress:

• electronic bibliographic databases

• contact with experts in the field

• references of included studies

• screening of relevant websites.

The following databases of published studies were searched: MEDLINE, MEDLINE In-Process & Other Non-Indexed Citations, EMBASE, Science Citation Index and Conference Proceedings CitationIndex – Science, The Cochrane Library [specifically the Cochrane Database of Systematic Reviews (CDSR), Cochrane Central Register of Controlled Trials (CENTRAL), Database of Abstracts of Reviews of Effects (DARE), NHS Economic Evaluation Database (NHS EED), HTA database], Current Controlled Trials, ClinicalTrials.gov and UK Clinical Research Network (UKCRN) Portfolio Database. The search strategies were initially developed for MEDLINE and were adapted as appropriate for other databases.

The reference lists of included studies and relevant review articles were checked and the following websites of hip implant manufacturers were screened for relevant publications:

• Amplitude

• Biomet

• B Braun/Aesculap

• Comis Orthopaedics

• Corin

• DePuy

• Exactech

• Finsbury

• JRI Orthopaedics

• Implantcast

• Implants International

• Lima WG Healthcare

• Mathys Orthopaedics

• Medacta UK

• Orthodynamics

• Peter Brehm

• SERF Dedienne santé

• Smith & Nephew

• Stanmore Implants Worldwide

• Stryker

• Symbios SA

• Waldemar Link

• Wright Medical UK

• Zimmer, Inc.

Grey literature searches were undertaken using Google (Google Inc., Mountain view, CA, USA) and the online resources of the following regulatory bodies, health services, research agencies and professional societies:

• British Hip Society

• British Orthopaedic Association

• Orthopaedic Research UK

• ODEP

• NJR

• Arthritis Research UK

• Cochrane Musculoskeletal Group

• Arthritis Care

• MHRA

• American Association of Hip and Knee Surgeons

• American Academy of Orthopedic Surgeons (AAOS)

• The Hip Society

• Royal College of Surgeons

• Royal College of Surgeons of Edinburgh.

All bibliographic records identified through the electronic searches were collected in a managed reference database.

Study design

• RCTs.

• Systematic reviews.

• Meta-analyses.

Given the wide scope and large amount of identified evidence, we limited studies to those published since 2008 with a sample size of ≥ 100 participants.

Population

• People with pain or disability resulting from end-stage arthritis of the hip for whom non-surgical management has failed.

Intervention

• Elective primary THR.

• Primary hip RS.

Comparator

• Different types of primary THR compared with RS for people in whom both procedures are suitable.

• Different types of primary THR compared with each other for people who are not suitable for hip RS.

Outcomes

Clinical effectiveness outcome measures were mortality, validated functional/pain and health-related quality of life total scores, revision rate, implant survival rate and femoral head penetration rate (measure of prosthesis movement). Adverse events included incidence of peri-/postprocedural complications (i.e. implant dislocation, infection, osteolysis, aseptic loosening, femoral fracture and deep-vein thrombosis).

Study and participant characteristics

Risk of bias and methodological quality

Risk of bias in the randomised controlled trials

The risk-of-bias assessments for the 13 included RCTs comparing different types of THR are presented in risk-of-bias tables (see Appendix 2), the summary table (Table 11) and the risk-of-bias graph (Figure 10). Overall, four112,119,123,128 of the 13 RCTs reported an adequate method for random sequence generation and eight110,112,119,123–126,129 reported adequate treatment allocation concealment (low risk of bias). A greater proportion of the RCTs were rated as having a low risk of performance and detection bias for objective (e.g. mortality, dislocation) than for subjective (e.g. patient-administered functional scores) outcomes (92–100% vs. 15–23%, respectively). For at least eight of the RCTs, it was unclear whether or not awareness of THR type would influence the ascertainment of clinical/functional scores by patients/study personnel (performance bias)110,112,113,115,124,125,127–129,145 or outcome assessors (detection bias). 112,113,115,124–126,128,129 Most RCTs failed to report the blinding status of the patients, study personnel and/or outcome assessors. Eight RCTs were judged as having a low risk of attrition bias. Five RCTs115,124,125,127,128 were judged as being at high risk for selective outcome and/or analysis bias. The risk of other biases (e.g. funding source, baseline imbalance in important characteristics, inappropriate analysis) for about one-third of the RCTs was judged to be high.

TABLE 11 - Risk of bias summary for RCTs: review authors’ judgements about each risk of bias item – THR vs. THR

–, high risk of bias; +, low risk of bias; ?, unclear risk of bias; ITT, intention to treat; NA, not applicable; PP, per protocol.

Study	Selection bias: random sequence generation	Selection bias: allocation concealment	Performance bias: subjective (e.g. patient reported)	Performance bias: objective (e.g. mortality, radiography, dislocation)	Detection bias: subjective (e.g. patient reported)	Detection bias: objective (e.g. mortality, radiography, dislocation)	Attrition bias: subjective (e.g. patient reported)	Attrition bias: objective (e.g. mortality, radiography, dislocation)	Reporting bias: selective reporting of the outcome, subgroups or analysis	Other bias [funding source, adequacy of statistical methods used, type of analysis (ITT/PP), baseline imbalance in important characteristics]
Amanatullah 2011125	?	+	?	+	?	+	?	–	–	–
Angadi 2012112	+	+	?	+	?	+	+	+	+	?
Bjørgul 2010110	?	+	?	+	+	+	–	–	+	–
Capello 2008115	?	?	?	+	?	+	+	+	–	–
Corten 2011119	+	+	+	+	+	+	+	+	+	+
Engh 2012113	?	?	?	?	?	+	–	–	+	?
Healy 2009127	–	–	?	+	–	+	+	+	–	+
Howie 2012123	+	+	NA	+	NA	+	NA	+	+	+
Kadar 2011126	?	+	+	+	+	+	+	+	+	+
Kim 2011128	+	?	?	+	?	+	+	+	–	+
Kim 2011129	?	+	?	+	?	+	+	+	+	?
Lewis 2008124	?	+	?	+	?	+	?	?	–	?
McCalden 2009145	?	?	?	+	+	+	+	+	+	–

FIGURE 10.

Risk of bias graph for RCTs: review authors’ judgements about each risk of bias item – THR vs. THR. ITT, intention to treat; NA, not applicable; PP, per protocol.

Clinical effectiveness findings for the comparison of different types of total hip replacement

This section summarises the evidence from the 13 RCTs110,112,113,115,119,123–129,145 and five systematic reviews. 137–141

The reported outcomes for this section were HHS (12 RCTs;110,112,113,115,119,124–129,145 three systematic reviews137,138,140), WOMAC score (four RCTs119,124,129,145), McMaster Toronto Arthritis Patient Preference Questionnaire (MACTAR) score (one RCT119), Merle d’Aubigné and Postel hip score (one RCT119), University of California Los Angeles (UCLA) activity score (one RCT129), OHS (one systematic review137), SF-12 score (three RCTs;124,125,145 one systematic review140), risk of revision (10 RCTs;112,113,115,119,123–125,127–129 five systematic reviews137–141), mortality (six RCTs110,113,119,123,128,145), femoral head penetration rate (three RCTs113,126,145), implant dislocation (seven RCTs;110,112,115,123–125,127 two systematic reviews139,140), osteolysis (seven RCTs;112,113,115,125,127,129,145 two systematic reviews138,139), aseptic loosening (five RCTs;112,113,119,124,127 one systematic review139), femoral fracture (three RCTs113,115,127), infection (four RCTs112,124,125,127) and deep-vein thrombosis (one RCT125).

Neither the RCTs nor the systematic reviews reported any evidence for the following clinical effectiveness outcomes:

• HOOS

• LISOH

• AAOS Hip and Knee Questionnaire

• Arthritis Impact Measurement Scale (AIMS)

• Nottingham Health Profile (NHP) questionnaire

• EQ-5D

• SF-36

• time to revision

• pain score [visual analogue scale (VAS)].

Summary results for the following outcomes are presented separately for RCTs and systematic reviews in the following sections. The outcomes of interest are as follows:

• mortality

• validated functional/pain (total scores): HHS, OHS, pain score (VAS), Merle d’Aubigné and Postel score, UCLA activity score, WOMAC, MACTAR, HOOS, LISOH, AAOS Hip and Knee Questionnaire, AIMS

• health-related quality of life (total scores): EQ-5D, SF-36/SF-12, NHP

• revision rate (risk of revision, mean time to revision)

• femoral head penetration rate (measure of prosthesis movement)

• adverse events (peri-/postprocedural complications): implant dislocation, infection, osteolysis, aseptic loosening, femoral fracture, deep-vein thrombosis, muscle weakness, nerve palsy and pulmonary embolism.

Functional/clinical measures

Twelve of the 13 included RCTs comparing different types of THR reported at least some results for the following functional scores measured at different postprocedure follow-up times: HHS (12 studies110,112,113,115,119,124–129,145) WOMAC score (four studies119,124,129,145), MACTAR score (one study119), Merle d’Aubigné and Postel score (one study119) and UCLA activity score (one study129). None of these 12 studies reported measurements of the OHS.

Three of the five included systematic reviews comparing different types of THR reported at least some evidence on HHS137,138,140 and OHS. 137 None of the three reviews reported any summary evidence for WOMAC, MACTAR, Merle d’Aubigné and Postel, and UCLA scores.

Harris Hip Score

Randomised controlled trials (n = 12)

Mean HHS at follow-up (range 6 months–10 years) did not differ between the following interventions: cup fixation (two studies;110,112 cemented vs. cementless), cup liner bearing surface (two studies;113,145 cross-linked polyethylene vs. non-cross-linked polyethylene), cup and femoral stem fixation (one study;119 cemented vs. cementless) and femoral head-on-cup liner bearing surfaces (one study;126 cobalt–chromium/oxinium-on-polyethylene vs. cobalt–chromium/oxinium-on-cross-linked polyethylene) (Table 13). The pooled MD for HHS in our meta-analysis of two studies113,145 comparing cup liners made with cross-linked polyethylene compared with non-cross-linked polyethylene was 2.29 (95% CI –0.88 to 5.45), suggesting a non-significant benefit of cross-linked polyethylene cup liners (Figure 11).

TABLE 13 - Harris Hip Score (range 0–100): RCTs

ANOVA, analysis of variance; CoCr, cobalt–chromium; HA, hydroxyapatite; NR, not reported; NS, not significant; PE, polyethylene; SROB, summary risk of bias; SS, statistically significant; XLPE, cross-linked polyethylene.

Decision was consensus based.

Favours THR-1 (or THR-2), no difference or inconclusive.

Calculated.

Only those comparisons listed for which any evidence for the given outcome was reported; one RCT did not report any evidence for this outcome.

Follow-up	Arm-specific estimate, mean (SD or 95% CI)	Difference (p-value or 95% CI)	Number of RCTs (SROB across studies)a	Treatment effect conclusionb
Cup fixation: cemented vs. cementless
6 months	90.2 (87.9 to 92.6) vs. 89.1 (86.9 to 91.3)110	p > 0.05 (NS)	2 (unclear)	No difference
2 years	92.7 (89.6 to 95.8) vs. 94.0 (92.4 to 95.7)110	p > 0.05 (NS)
5 years	93.9 (91.6 to 96.2) vs. 91.4 (89.3 to 93.5)110	p > 0.05 (NS)
10 years	89.8 (87.0 to 92.6) vs. 87.3 (84.1 to 90.6)110	p > 0.05 (NS)
10 years	74.5 (NR) vs. 78.0 (NR)112	p > 0.05 (NS)
Cup liner bearing surface: XLPE vs. non-XLPE
1 year	85.0 (10.3) vs. 83.4 (13.1)145	MD 1.60 (–3.07 to 6.27)c	2 (unclear)	No difference
5 years	86.0 (13.1) vs. 83.1 (15.4)145	MD 2.90 (–2.77 to 8.57)c
10 years	88.0 (14.0) vs. 86.0 (15.0)113	MD 2.00 (–1.85 to 5.85)c
		Pooled estimate of MDc 2.29 (–0.88 to 5.45)113,145
Cup shell design: porous-coated shell vs. arc-deposited HA-coated shell
5 years	97.0 (NR) vs. 96.4 (NR)115	p > 0.05 (NS)	1 (unclear)	Inconclusive
10 years	96.0 (NR) vs. 96.7 (NR)115	p > 0.05 (NS)
Cup and femoral stem fixation: cemented cup/femoral stem vs. cementless cup/femoral stem
3 months	41 (12.0) vs. 41 (11.0)119	MD 0.0 (–3.00 to 3.00)c	1 (low)	No difference
6 months	47 (12) vs. 50 (13)119	MD –3.0 (–6.32 to 0.32)c
1 year	52 (10.0) vs. 53 (11.0)119	MD –1.0 (–3.86 to 1.86)c
3 years	50 (14.0) vs. 52 (11.0)119	MD –2.0 (–5.62 to 1.62)c
5 years	47 (14.0) vs. 48 (13.0)119	MD –1.0 (–4.88 to 2.87)c
7 years	44 (15) vs. 46 (14)119	MD –2.0 (–7.07 to 3.05)c
Femoral head bearing surface: oxinium femoral heads vs. CoCr femoral heads
2 years	92 (NR) vs. 92.5 (NR)124	p > 0.159 (NS)	1 (unclear)	Inconclusive
Femoral head-on-cup liner bearing surface: ceramic-on-ceramic vs. metal-on-PE
5 years	96.4 (NR) vs. 97.0 (NR)115	p > 0.05 (NS)	1 (unclear)	Inconclusive
10 years	96.7 (NR) vs. 96.4 (NR)115	p > 0.05 (NS)
Femoral head-on-cup liner bearing surface: ceramic-on-ceramic vs. ceramic-on-PE
5 years	NR125	p > 0.05 (NS)	1 (unclear)	Inconclusive
Femoral head-on-cup liner bearing surface: steel-on-PE vs. CoCr-on-PE vs. oxinium-on-PE vs. CoCr-on-XLPE vs. oxinium-on-XLPE
2 years	91 (10.8) vs. 91 (8.5) vs. 91 (11.1) vs. 93 (11.3) vs. 88 (9.5)126	p = 0.7 (NS); ANOVA-based p = 0.5 (NS)c	1 (low)	No difference
Femoral stem composition: CoCr vs. titanium
5 years	83 (NR) vs. 87 (NR)127	p = 0.029 (SS)	1 (high)	Inconclusive
Femoral stem design: short metaphyseal-fitting stem vs. conventional metaphyseal- and diaphyseal-fitting stem
3 years	97.0 (NR) vs. 96.0 (NR)128	p = 0.79 (NS)	1 (unclear)	Inconclusive
Femoral stem fixation: cemented vs. cementless
18 years	91 (NR) vs. 90 (NR)129	p = 0.71 (NS)	1 (unclear)	Inconclusive

FIGURE 11.

Harris Hip Score. XLPE, cross-linked polyethylene.

The evidence for the other comparisons based on cup shell design (porous coated vs. arc-deposited hydroxyapatite coated),115 femoral head bearing surface (oxinium vs. cobalt–chromium),124 femoral head-on-cup liner bearing surfaces (ceramic-on-ceramic vs. metal-on-polyethylene or ceramic-on-polyethylene),115,125 femoral stem composition (cobalt–chromium vs. titanium),127 femoral stem design (short metaphyseal fitting vs. conventional diaphyseal fitting)128 and femoral stem fixation (cemented vs. cementless)129 was judged to be inconclusive.

Systematic reviews (n = 3)

One systematic review140 reported the pooled MD for the HHS (Table 14). Pooled estimates for the comparison between metal-on-metal and metal-on-polyethylene bearing surfaces at two different follow-up times were not consistent: at 2 years metal-on-metal bearing surfaces gave a significantly higher HHS than metal-on-polyethylene, but at > 2 years there was no significant difference between the two types of THR. The remaining two systematic reviews presented only narrative summaries. 137,138 In summary, for the HHS the systematic review-based evidence was judged to be inconclusive.

TABLE 14 - Harris Hip Score (range 0–100): systematic reviews

MA, meta-analysis; NR, not reported; NS, not significant; PE, polyethylene; SS, statistically significant.

Favours THR-1 (or THR-2), no difference or inconclusive.

Only those comparisons listed for which any evidence for the given outcome was reported. Two systematic reviews did not report this outcome.

Follow-up	Pooled effect estimate (95% CI)	Number of RCTs in MA or narrative synthesis	AMSTAR rating	Treatment effect conclusiona
Cup fixation: cemented vs. cementless
3 years	NR137	2137	High quality137	Inconclusive
2–5 years	NR138	3138	Low quality138
Femoral head-on-cup liner surface: metal-on-metal vs. metal-on-PE
2 years	MD –2.40 (–4.47 to –0.33) (SS)140	4140	High quality140	Inconclusive
> 2 years	MD 1.21 (–2.41 to 4.83) (NS)140	2140
Femoral head-on-cup liner surface: ceramic-on-ceramic vs. ceramic-on-PE
NR	NR140	5140	High quality140	Inconclusive
Femoral head-on-cup liner surface: ceramic-on-PE vs. metal-on-PE
NR	NR140	2140	High quality140	Inconclusive
Femoral head-on-cup liner surface: metal-on-metal vs. ceramic-on-ceramic
NR	NR140	1140	High quality140	Inconclusive

Western Ontario and McMaster University Osteoarthritis Index score

RCTs (n = 4)

Results from all four RCTs reporting postprocedural mean WOMAC scores indicated statistically non-significant differences between the THR groups compared with respect to cup liner bearing surface (cross-linked polyethylene vs. non-cross-linked polyethylene),145 cup and femoral stem fixation (cemented vs. cementless),119 femoral head bearing surface (oxinium vs. cobalt–chromium)124 and femoral stem fixation (cemented vs. cementless)129 (Table 15). The MD in WOMAC score of –0.12 (95% CI –7.58 to 7.34) observed for one RCT145 suggested no difference between cross-linked polyethylene and non-cross-linked polyethylene cup liners. Results for WOMAC score in the remaining three RCTs119,124,129 were judged to be inconclusive because of incompletely reported data.

TABLE 15 - Western Ontario and McMaster University Osteoarthritis Index (range 0–100): RCTs

CoCr, cobalt–chromium; NA, not applicable; NR, not reported; NS, not significant; SROB, summary risk of bias; XLPE, cross-linked polyethylene.

Decision was consensus based.

Favours THR-1 (or THR-2), no difference or inconclusive.

Calculated.

Only those comparisons listed for which any evidence for the given outcome was reported. Nine RCTs110–118,123,125–128 did not report any evidence on this outcome.

Follow-up	Arm-specific estimate, mean (SD or 95% CI)	Difference (p-value or 95% CI)	Number of RCTs (SROB)a	Treatment effect conclusionb
Cup liner bearing surface: XLPE vs. non-XLPE
1 year	83.0 (17.2) vs. 81.6 (17.6)145	MD 1.43 (–5.48 to 8.34)c	1 (unclear)	No difference
5 years	78.0 (19.4) vs. 78.1 (18.2)145	MD –0.12 (–7.58 to 7.34)c
Cup and femoral stem fixation: cemented cup/femoral stem vs. cementless cup/femoral stem
NA	Mean domain subscores only119	–	1 (low)	NA
Femoral head bearing surface: oxinium femoral heads vs. CoCr femoral heads
2 years	84.9 (NR) vs. 87.0 (NR)124	p > 0.159 (NS)	1 (unclear)	Inconclusive
Femoral stem fixation: cemented vs. cementless
16 years	11 (NR) vs. 13 (NR)129	p = 0.927 (NS)	1 (unclear)	Inconclusive

Systematic reviews (n = 0)

No evidence was identified.

Other functional/clinical scores

Randomised controlled trials (n = 2)

In one RCT119 there was no difference in mean MACTAR scores (at 7 years: mean change difference 0.20, 95% CI –0.74 to 1.14) and Merle d’Aubigné and Postel scores (at 7 years: mean change difference –0.40, 95% CI –1.34 to 0.54) between patients who received a THR with cemented components and those who received a THR with cementless components (Tables 16 and 17). Results from one RCT129 comparing femoral stem fixation (cemented vs. cementless) by the postoperative UCLA activity score were inconclusive because of incomplete data reporting (Table 18).

TABLE 16 - McMaster Toronto Arthritis Patient Preference Questionnaire scores (range 0–30): RCTs

SROB, summary risk of bias.

Decision was consensus based.

Favours THR-1 (or THR-2), no difference or inconclusive.

Calculated.

Only those comparisons listed for which any evidence for the given outcome was reported. Only the study by Corten et al. 119 reported any evidence for this outcome.

Follow-up	Arm-specific estimate, mean (SD)	Difference (95% CI)	No. of RCTs (SROB)a	Treatment effect conclusionb
Cup and femoral stem fixation: cemented cup/femoral stem vs. cementless cup/femoral stem
	Mean change (postoperative):	Mean change difference:	1 (low)	No difference
3 months	–5.3 (2.5) vs. –5.2 (2.2)119	MD 0.10 (–0.51 to 0.71)c
6 months	–6.6 (1.9) vs. –6.4 (2.1)119	MD 0.20 (–0.33 to 0.73)c
1 year	–7.0 (1.8) vs. –6.9 (2.0)119	MD 0.10 (–0.41 to 0.61)c
3 years	–6.6 (2.3) vs. –6.4 (2.3)119	MD 0.20 (–0.46 to 0.86)c
5 years	–6.0 (2.8) vs. –6.2 (2.4)119	MD –0.20 (–0.45 to 0.55)c
7 years	–6.2 (2.8) vs. –6.0 (2.6)119	MD 0.20 (–0.74 to 1.14)c

TABLE 17 - Merle d’Aubigné and Postel scores (range 0–18): RCTs

SROB, summary risk of bias.

Decision was consensus based.

Favours THR-1 (or THR-2), no difference or inconclusive.

Calculated.

Only those comparisons listed for which any evidence for the given outcome was reported. Only the study by Corten et al. 119 reported any evidence for this outcome.

Follow-up	Arm-specific estimate, mean (SD)	Difference (95% CI)	Number of RCTs (SROB)a	Treatment effect conclusionb
Cup and femoral stem fixation: cemented cup/femoral stem vs. cementless cup/femoral stem
	Mean change (postoperative):	Mean change difference:	1 (low)	No difference
3 months	5.8 (1.9) vs. 5.6 (2.2)119	MD 0.20 (–0.34 to 0.74)c
6 months	6.7 (2.1) vs. 7.0 (2.2)119	MD –0.30 (–0.87 to 0.27)c
1 year	7.5 (1.8) vs. 7.4 (2.1)119	MD 0.10 (–0.43 to 0.63)c
3 years	7.1 (2.2) vs. 6.9 (2.1)119	MD 0.20 (–0.41 to 0.81)c
5 years	6.5 (2.3) vs. 6.6 (2.4)119	MD –0.10 (–0.77 to 0.57)c
7 years	6.1 (2.6) vs. 6.5 (2.8)119	MD –0.40 (–1.34 to 0.54)c

TABLE 18 - University of California Los Angeles activity scores (range 1–10): RCTs

NR, not reported; NS, not significant; SROB, summary risk of bias.

Decision was consensus based.

Favours THR-1 (or THR-2), no difference or inconclusive.

Only those comparisons listed for which any evidence for the given outcome was reported. Only the study by Kim et al. 129 reported any evidence for this outcome.

Follow-up	Arm-specific estimate, mean (SD or 95% CI)	Difference (p-value)	Number of RCTs (SROB)a	Treatment effect conclusionb
Femoral stem fixation: cemented vs. cementless
16 years	7.6 (NR) vs. 7.8 (NR)129	p = 0.814 (NS)	1 (unclear)	Inconclusive

Systematic reviews (n = 1)

The OHS was reported in one systematic review137 comparing cup fixation methods (cemented vs. cementless), but the results were inconclusive (Table 19). This evidence was based on one RCT showing a statistically non-significant result.

TABLE 19 - Oxford Hip Score (range 0–48): systematic review

MA, meta-analysis; NR, not reported.

Favours THR-1 (or THR-2), no difference or inconclusive.

Only those comparisons listed for which any evidence for the given outcome was reported. Only the study by Voigt et al. 137 reported this outcome.

Follow-up	Pooled effect estimate (95% CI)	Number of RCTs in MA or narrative synthesis	AMSTAR rating	Treatment effect conclusiona
Cup fixation: cemented vs. cementless
3 years	NR137	1137	High quality137	Inconclusive

Mortality

Evidence on mortality was reported for six RCTs. 110,113,119,123,128,145 None of the five systematic reviews reported on mortality.

Randomised controlled trials (n = 6)

Evidence from the six RCTs110,113,119,123,128,145 that reported mortality was inconclusive because of non-significant RR estimates and wide 95% CIs (Table 24). For example, based on a pooled RR estimate of 1.39 (95% CI 0.78 to 2.49),113,145 5- to 10-year post-surgery mortality rates in the group receiving cross-linked polyethylene cup liners were not significantly different from those in the group receiving non-cross-linked polyethylene cup liners (Figure 12). Similarly, the rest of the studies showed non-significant results for mortality between THR groups defined by femoral stem and/or cup fixation (cemented vs. cementless)110,119 and femoral head size (36 mm vs. 28 mm). 123 One RCT128 reported no deaths for both treatment groups receiving femoral stems of different design.

TABLE 24 - Mortality rate: RCTs

MH-RR, Mantel–Haenszel rate ratio; NR, not reported; PE, polyethylene; NS, not significant; SROB, summary risk of bias; XLPE, cross-linked polyethylene.

Decision was consensus based.

Favours THR-1 (or THR-2), no difference or inconclusive.

Calculated.

Only those comparisons listed for which any evidence for the given outcome was reported. Seven RCTs112,115–118,124–127,129 did not report any evidence for this outcome.

Follow-up	Arm-specific count (n/N)	Difference (p-value or 95% CI)	Number of RCTs (SROB across studies)a	Treatment effect conclusionb
Cup fixation: cemented vs. cementless
10 years	12/107 vs. 14/108110	p = NR; RR 0.86 (95% CI 0.41 to 1.78)c	1 (low)	Inconclusive
Cup liner bearing surface: XLPE vs. non-XLPE
5 years	7/50 vs. 2/50145	p > 0.05 (NS); RR 3.50 (95% CI 0.76 to 16.03)c	2 (unclear)	Inconclusive
10 years	17/111 vs. 15/109113	p > 0.05 (NS); RR 1.11 (95% CI 0.58 to 2.11)c
		Pooled estimate of MH-RR: RR 1.39 (95% CI 0.78 to 2.49)113,145
Cup and femoral stem fixation: cemented cup/femoral stem vs. cementless cup/femoral stem
7 years	18/124 vs. 17/126119	p = NR; RR 1.07 (95% CI 0.58 to 1.98)c	1 (low)	Inconclusive
Femoral head size: 36 mm vs. 28 mm
1 year	5/273 vs. 2/284123	p = NR; RR 2.58 (95% CI 0.53 to 13.20)c	1 (low)	Inconclusive
Femoral stem design: short metaphyseal-fitting stem vs. conventional metaphyseal- and diaphyseal-fitting stem
3 years	0/50 vs. 0/50128	p = NR; RR and 95% CI not estimated	1 (low)	Inconclusive

FIGURE 12.

Mortality. XLPE, cross-linked polyethylene.

Systematic reviews (n = 0)

No evidence was identified.

Implant dislocation

Randomised controlled trials (n = 7)

Evidence on the occurrence/absence of implant dislocation was reported by seven RCTs110,112,115,123–125,127 (Table 26). Our pooled estimate of two studies110,112 (Figure 13) indicated a reduced risk of implant dislocation at 10 years’ follow-up in recipients of cemented compared with cementless cups (pooled OR 0.34, 95% CI 0.13 to 0.89). Moreover, in one RCT123 after 1 year of follow-up, the THR recipients with a larger size of femoral head experienced a lower risk of implant dislocation than those with a smaller size of femoral head (36 mm vs. 28 mm: RR 0.17, 95% CI 0.04 to 0.78). Evidence on implant dislocation for the remaining four RCTs115,124,125,127 was inconclusive because of incomplete data and non-significant results.

TABLE 26 - Implant dislocation rate: RCTs

CoCr, cobalt–chromium; HA, hydroxyapatite; NR, not reported; NS, not significant; PE, polyethylene; SROB, summary risk of bias.

Decision was consensus based.

Favours THR-1 (or THR-2), no difference or inconclusive.

Calculated.

Only those comparisons listed for which any evidence for the given outcome was reported. Six RCTs113,119,126,128,129,145 did not report any evidence for this outcome.

Follow-up	Arm-specific count, n/N	Difference (p-value or 95% CI)	Number of RCTs (SROB across studies)a	Treatment effect conclusionb
Cup fixation: cemented vs. cementless
10 years	4/107 vs. 10/108110	p > 0.05 (NS); RR 0.40 (95% 0.13 to 1.24)c	2 (low)	In favour of cemented cup
	1/183 vs. 3/104112	p = NR; RR 0.18 (95% 0.02 to 1.79)c
		Pooled estimate of Peto OR: 0.34 (95% 0.13 to 0.89)110,112c
Cup shell design: porous-coated shell vs. arc-deposited HA-coated shell
10 years	2/113 vs. 3/109115	p = NR; RR 0.64 (95% 0.10 to 3.77)c	1 (low)	Inconclusive
Femoral head size: 36 mm vs. 28 mm
1 year	2/258 vs. 12/275123	p = NR; RR 0.17 (95% 0.04 to 0.78)c	1 (low)	In favour of 36-mm head size
Femoral head bearing surface: oxinium femoral heads vs. CoCr femoral heads
2 years	2/50 vs. 1/50124	p = NR; RR 2.00 (95% 0.18 to 21.35)c	1 (low)	Inconclusive
Femoral head-on-cup liner bearing surface: ceramic-on-ceramic vs. ceramic-on-PE
5 years	10/166 vs. 9/146125	p = 0.672 (NS); RR 0.97 (95% CI 0.40 to 2.33)c	1 (low)	Inconclusive
Femoral head-on-cup liner bearing surface: ceramic-on-ceramic vs. metal-on-PE
10 years	5/222 vs. 5/106115	p = 0.25 (NS); RR 0.47 (95% CI 0.14 to 1.61)c	1 (low)	Inconclusive
Femoral stem composition: CoCr vs. titanium
5 years	3/199 vs. 0/191127	p = 0.678 (NS); RR and 95% CI not estimated	1 (unclear)	Inconclusive

FIGURE 13.

Implant dislocation.

Systematic reviews (n = 2)

Overall, no conclusions on implant dislocation could be drawn from the two systematic reviews, given the narrative evidence summary140 and the mixed study designs139 (Table 27). The pooled data from one review139 was based on nine studies, most of which were not randomised and which indicated a lower risk of dislocation in the groups receiving cemented compared with cementless cups.

TABLE 27 - Implant dislocation rate: systematic reviews

MA, meta-analysis; NR, not reported; PE, polyethylene.

Favours THR-1 (or THR-2), no difference or inconclusive.

Only those comparisons listed for which any evidence for the given outcome was reported.

Follow-up	Pooled effect estimate (95% CI)	Number of RCTs in MA or narrative synthesis	AMSTAR rating	Treatment effect conclusiona
Cup fixation: cemented vs. cementless
5–15 years	12/914 (1.3%) vs. 28/696 (4.1%) (p = 0.001).139 Pooled data from nine comparative studies (most non-RCTs) suggested that cemented cups had lower dislocation rates than cementless cups	NR139	Low quality139	Inconclusive
Femoral head-on-cup liner surface: metal-on-metal vs. metal-on-PE
2–5 years	NR.140 No significant difference based on results from three RCTs	3140	High quality140	Inconclusive

Summary conclusions for the comparison between different types of total hip replacement

Other analysis

Study and participant characteristics

Risk of bias and methodological quality

Risk of bias in randomised controlled trials

The risk of bias assessment for the three included RCTs130–132 comparing THR with RS is presented in risk of bias tables (see Appendix 2), the summary table (Table 39) and the risk of bias graph (Figure 14). Overall, two studies130,132 reported an adequate method for random sequence generation and all three studies130–132 reported treatment allocation concealment (low risk of bias). Two of the three studies130,132 were rated as having a low risk of performance and detection bias for objective outcomes (e.g. revision, dislocation). The same two studies had a high risk of performance bias for subjective outcomes (e.g. patient-administered functional scores). Patients and study personnel were blinded in only one study. 131 For two studies130,132 the influence of attrition bias on objective outcomes was judged to be of low risk. All three studies were judged as being at low risk for selective outcome and/or analysis bias. Risk of other biases (e.g. funding source, balance/imbalance in important characteristics, inappropriate analysis) for one of the three studies was judged to be high. 131

TABLE 39 - Risk of bias summary for RCTs: review authors’ judgements about each risk of bias item – THR vs. RS

–, high risk of bias; +, low risk of bias; ?, unclear risk of bias; ITT, intention to treat; NA, not applicable; PP, per protocol.

Study	Selection bias: random sequence generation	Selection bias: allocation concealment	Performance bias: subjective (e.g. patient reported)	Performance bias: objective (e.g. mortality, radiography, dislocation)	Detection bias: subjective (e.g. patient reported)	Detection bias: objective (e.g. mortality, radiography, dislocation)	Attrition bias: subjective (e.g. patient reported)	Attrition bias: objective (e.g. mortality, radiography, dislocation)	Reporting bias: selective reporting of the outcome, subgroups or analysis	Other bias [funding source, adequacy of statistical methods used, type of analysis (ITT/PP), baseline imbalance in important characteristics]
Costa 2012130	+	+	–	+	+	+	+	+	+	+
Garbuz 2010131	?	+	+	NA	?	NA	+	NA	+	–
Vendittoli 2010132	+	+	–	+	?	+	–	+	+	+

FIGURE 14.

Risk of bias graph for RCTs: review authors’ judgements about each risk of bias item – THR vs. RS. ITT, intention to treat; NA, not applicable; PP, per protocol.

Clinical effectiveness findings for the comparison between total hip replacement and resurfacing arthroplasty

This section summarises the findings from the three RCTs130–132 and three systematic reviews. 142–144

The reported outcomes for this section were the HHS (one RCT;130 two systematic reviews142,143), WOMAC score (two RCTs;131,132 two systematic reviews142,143), Merle d’Aubigné and Postel score (one RCT;132 one systematic review142), UCLA activity score (two RCTs;131,132 one systematic review42), OHS (one RCT130), health-related quality of life scales (SF-36 and EQ-5D; two RCTs130,131), risk of revision (one RCT;132 two systematic reviews142,143), mortality (two systematic reviews142,143), infection (two RCTs;130,132 one systematic review142), aseptic loosening (one RCT;132 two systematic reviews142,143), implant dislocation (two RCTs;130,132 one systematic review142) and deep-vein thrombosis (two RCTs130,132).

Neither the RCTs nor the systematic reviews reported any evidence for the following clinical effectiveness outcomes:

• HOOS

• LISOH

• AAOS Hip and Knee Questionnaire

• AIMS

• MACTAR

• NHP questionnaire

• SF-12

• time to revision

• pain score (VAS)

• femoral head penetration.

Summary results for the included outcomes are presented separately for RCTs and systematic reviews.

Evidence from randomised controlled trials

Complications

Evidence on complications was reported for two RCTs. 130,132 Meta-analysis of the data on risk of infection from the two RCTs indicated that, at 12–56 months post operation, THR recipients were at an increased risk of infection compared with RS recipients (pooled OR 7.94, 95% CI 1.78 to 35.40) (Figure 15). In addition, evidence on the differences between groups for the risk of deep-vein thrombosis (Figure 16; pooled OR 0.60, 95% CI 0.15 to 2.42),130,132 implant dislocation (Figure 17; pooled OR 3.97, 95% CI 0.79 to 19.90),130,132 wound complications (RR 4.01, 95% CI 0.92 to 18.18)130 and aseptic loosening (RR not estimable)132 was judged to be inconclusive by us.

FIGURE 15.

Risk of infection.

FIGURE 16.

Risk of deep-vein thrombosis.

FIGURE 17.

Risk of implant dislocation.

A summary of the results for the difference outcomes is presented in Table 41.

TABLE 41 - Summary of the results for THR vs. RS: RCTs

MCS, mental component summary score; NR, not reported; NS, not significant; PCS, physical component summary score; SROB, summary risk of bias; SS, statistically significant.

Decision was consensus-based.

Favours THR (or RS), no difference or inconclusive.

Mean change in WOMAC score from baseline.

Calculated.

End-point mean WOMAC score.

Only those outcomes listed for which any evidence was reported; studies not reporting a given outcome are not listed.

Follow-up (months)	Arm-specific estimates, n/N or mean (SD or 95% CI) (THR vs. RS)	Difference (p-value or 95% CI)	Number of RCTs (SROB across studies)a	Treatment effect conclusionb
HHS (range 0–100)
12	82.3 (77.2 to 87.5) vs. 88.4 (84.4 to 92.4)130	MD –6.04 (12.58 to 0.51)	1 (low)	Inconclusive
OHS (range 0–48)
12	38.2 (35.3 to 41.0) vs. 40.4 (37.9 to 42.9)130	MD –2.23 (–5.98 to 1.52)	1 (low)	No difference
WOMAC score (range 0–100)
3	c19.2 (NR) vs. 19.9 (NR)132	p = 0.76 (NS)132,136	2 (unclear)	No difference
6	c11.3 (NR) vs. 13.9 (NR)132	p = 0.20 (NS)132,136
12	c10.2 (10.7) vs. 8.0 (13.2)132	dMD 2.20 (–1.57 to 5.97)132,136
12	e90.18 (NR) vs. 90.40 (NR)131	p = 0.95 (NS)131
24	c9.0 (11.9) vs. 5.7 (8.6)132	dMD 3.30 (0.01 to 6.58)132,136
Merle d’Aubigné and Postel score (range 0–18)
3	15.8 (NR) vs. 16.2 (NR)132	p = 0.59 (NS)	1 (unclear)	No difference
6	17.1 (NR) vs. 17.2 (NR)132	p = 0.72 (NS)
12	16.6 (NR) vs. 16.7 (NR)132	p = 0.94 (NS)
24	17.5 (1.3) vs. 17.5 (1.3)132	p = 0.94 (NS); MD 0.0 (–1.06 to 1.06)c
UCLA activity score (range 1–10)
12	6.3 (NR) vs. 6.8 (NR)131	p = 0.24 (NS)131	2 (unclear)	Inconclusive
12	6.3 (NR) vs. 7.1 (NR)132	p = 0.03 (SS)132,136
24	NR (NR) vs. NR (NR)132	p = 0.09 (NS)132,136
SF-36 score (range 0–100)
12	MCS: 55.13 (NR) vs. 53.87 (NR)131	p = 0.55 (NS)	1 (unclear)	Inconclusive
12	PCS: 51.28 (NR) vs. 51.22 (NR)131	p = 0.97 (NS)
EQ-5D score (range 0–1)
12	0.71 (0.63 to 0.80) vs. 0.79 (0.72 to 0.87)130	MD –0.077 (–0.188 to 0.034)	1 (low)	Inconclusive
Revision rate
3	1/102 vs. 0/103132	p = NR; RR and 95% CI not estimated	1 (low)	Inconclusive
6	1/102 vs. 1/103132	p = NR; RR 1.01 (0.06 to 15.92)d
12	1/102 vs. 2/103132	p = NR; RR 0.50 (0.04 to 5.48)d
24	1/102 vs. 2/103132	p = NR; RR 0.50 (0.04 to 5.48)d
56	2/100 vs. 4/109132	p = 0.47 (NS); RR 0.54 (0.10 to 2.91)d
Complications
Infection
12	2/66 vs. 0/60130	p = 0.49 (NS); RR and 95% CI not estimated	2 (low)	In favour of RS
56	5/100 vs. 0/109132	p = 0.02 (SS); RR and 95% CI not estimated
		dPooled estimate of Peto OR: 7.94 (1.78 to 35.40)130,132
Deep-vein thrombosis
12	0/66 vs. 4/60130	p = 0.05 (NS); RR and 95% CI not estimated	2 (low)	Inconclusive
56	3/100 vs. 1/109132	p = NR (NS); RR 3.27 (95% CI 0.30 to 30.90)d
		dPooled estimate of Peto OR: 0.60 (95% CI 0.15 to 2.42)130,132
Implant dislocation
12	1/66 vs. 1/60130	p = 1.00 (NS); RR 0.90, 95% CI 0.05 to 14.21d	2 (low)	Inconclusive
56	4/100 vs. 0/109132	p = 0.038 (SS); RR and 95% CI not estimated
		dPooled estimate of Peto OR: 3.97 (95% CI 0.79 to 19.90)130,132
Superficial wound complication
12	9/66 vs. 2/60130	p = 0.06 (NS); RR 4.01 (95% CI 0.92 to 18.18)d	1 (low)	Inconclusive
Aseptic loosening
56	0/100 vs. 6/109132	p = 0.017 (SS); RR and 95% CI not estimated	1 (low)	Inconclusive

Evidence from systematic reviews

Summary conclusions for the comparison between total hip replacement and resurfacing arthroplasty

The majority of the evidence from three RCTs130–132 (Table 44) and three systematic reviews142–144 (Table 45) comparing THR and RS was rated as inconclusive (RCTs – very low-grade evidence). Nevertheless, the evidence from two RCTs and two systematic reviews indicated a reduced risk of infection130,132 and implant dislocation142,143 among RS patients compared with THR patients. However, the evidence from the same reviews also indicated that recipients of RS were at higher risk of revision and component loosening than patients who received a THR. In three RCTs130–132 the mean postoperative OHS, WOMAC score (low-grade evidence) and Merle d’Aubigné and Postel score were not different between patients who received THR and those who received RS.

There was inconclusive evidence on mortality (three RCTs130–132 and two systematic reviews142,143), HHS (one RCT130 and two systematic reviews142,143), UCLA activity score (two RCTs131,132 and one systematic review142) and selected complications (i.e. infection, wound complication, deep-vein thrombosis; two RCTs130,132 and one systematic review142).

Results from individual RCTs were considered inconclusive because of the partial reporting (missing data for effect estimates, CIs, SEs, SDs, p-values) and great uncertainty in the estimates (wide CIs). The findings from the systematic reviews were inconclusive because of great uncertainty in the pooled estimates (wide CIs), lack of reporting of pooled results across RCTs (i.e. only narrative synthesis reported) or inconsistent summary findings.

Other analysis

Inclusion criteria

Exclusion criteria

• Non-English-language publications.

• Abstract/conference proceedings, letters and commentaries.

• Quality of life reported without utilities or QALYs.

• Hip/knee data not reported separately.

• Studies including only patients aged < 35 years.

Resurfacing arthroplasty

Evidence on RS was scarce with only five of the 66 included studies investigating hip RS (see Appendix 10). A 2012 UK RCT including 126 OA patients suitable for RS investigated the cost-effectiveness of RS compared with THR. 40,130 At the end of this 12-month trial small benefits of RS in terms of QALYs could be shown for a selected patient group, resulting in an ICER of £17,451 per QALY. This evidence was stronger for male than for female patients. In a comparison between ceramic-on-ceramic THR and RS at 3 months post surgery, evidence was not as strong, favouring THR over RS. 208 However, longer-term follow-up in a study comparing hybrid THR with RS confirmed that, after 5 and 9 years, the revision rates for RS were lower than for hybrid THR (9.3% and 16.7% at 9 years post surgery, respectively) and patients were more active. 251,252

A retrospective economic decision analysis of published data over a 30-year time horizon showed the cost-effectiveness of RS compared with THR for women aged < 55 years and men aged < 65 years. 253 The main drivers of cost-effectiveness were the cost of the implant and length of hospital stay. 40,208 However, Vale et al. 19 reported in their HTA that RS would be cost-effective compared with THR only if RS revision rates could be shown to be 80–88% lower than revision rates for THR. They further concluded that RS could be cost-effective compared with ‘watchful waiting’ followed by THR or an extended period of ‘watchful waiting’ over 20 years.

Total hip replacement

The majority of studies investigated THR (n = 61) (see Appendix 10). Of these, five compared minimally invasive techniques with standard THR, reporting perioperative advantages, better short-term outcomes and reduced costs in favour of minimally invasive techniques. 11,148,254–256 However, Coyle et al. 148 concluded that there is little evidence of a difference between the two surgical techniques in the long term, mainly because of lack of data.

Ten of the THR studies focused on the comparison of different types of THR or specific components/brands of THR. Briggs et al. ,38 Davies et al. ,43 Fordham et al. 257 and Hulleberg et al. 258 assessed different brands of THR, Bozic et al. 259 investigated alternative bearings including metal-on metal, ceramic-on-ceramic and ceramic-on-polyethylene and Laupacis et al. ,120 Marinelli et al. ,260 Pennington et al. 44 and di Tanna et al. 261 compared cemented, cementless and hybrid THR more generally and reported inconsistent findings. The most recent economic model by Pennington et al. 44 used PROMs and showed that (1) cemented prostheses were the least costly type for THR, (2) hybrid prostheses were the most cost-effective and (3) cementless prostheses did not provide sufficient improvement in health outcomes to justify their additional costs. Similarly, Davies et al. 43 identified cemented prostheses as the least costly type of prosthesis in their review. However, they concluded that there is a lack of observed long-term prosthesis survival data and particularly limited up-to date evidence for the UK, which led them to call for more trials with longer-term follow-up. Cummins et al. 262 reported that use of antibiotic-impregnated bone cement can result in an overall decrease in costs. For more detail on the studies investigating the different types of THR see Appendix 12.

Patient management and rehabilitation was the focus of four studies,263–266 which reported that perioperative management and rehabilitation programmes could improve patient outcomes and reduce costs.

The majority of the THR studies (30/618,37,267–285,298,300–302,304–308) assessed the costs and/or effectiveness of THR without a specific focus on a rehabilitation programme, surgical intervention, implant brand or prosthesis type. Of these, two US studies267,268 concentrated on obese patients and reported that, even though operative costs are higher for obese patients, overall care costs and in-hospital outcomes for THR are comparable across all BMI groups. Eleven studies269–279 evaluated the cost-effectiveness of THR in a specific country, and two multicentre studies280,281 aimed to assess the costs and outcomes of THR comparatively across a number of European member states. These two studies concluded that improvement after surgery is associated with high preoperative expectations. Stargardt et al. 280 reported further that the total cost of treatment ranged from €1290 (Hungary) to €8739 (the Netherlands) and that the two main cost drivers were the cost of the implants and ward costs.

The overall findings of the cost-effectiveness studies were that (1) THR resulted in greater benefits than conservative treatment and (2) longer waiting times incurred greater costs and resulted in physical deterioration. 271,282,283 Further, agreement was reached on the long-term cost-effectiveness and sustained benefits of THR. 37,120,257,273,275 However, Bozic et al. 284 stated that, although THR improved quality of life, failed THR could lead to health states worse than chronic OA. Resource use might be increased as patients with a THR were shown to have a 10% increase in hospital stay compared with patients pre surgery. 285

In contrast, two studies286,287 that took a patient perspective rather than a health-care perspective concluded that out-of-pocket costs (including hospital costs, medication costs, rehabilitation costs, costs of health professional visits, costs of tests, costs of special equipment, costs of household alterations, use of private and community services and transportation costs that are not paid for by the health system), as well as use of health services, fell dramatically in the first year post surgery and that costs as well as resource use depended on pre-surgery health status.

Studies that focused on revision THR concluded that revision THR seems cost-effective but that it is resource intensive and has important implications for the allocation of health-care funding as the number of revisions is expected to increase with increasing demand for THR. 288–293 Vanhegan et al. 292 evaluated the costs associated with revision THR for different indications and reported that costs vary significantly by indication and that these variations were not reflected in the NHS tariffs. Durable implants and reduction in complications such as early dislocations have been suggested to be the solutions to reduce revision rates. 289 However, the highest revision costs were reported for revision as a result of infection,292 with infections caused by methicillin-resistant strains of bacteria (41% of periprosthetic joint infections) incurring significantly higher costs than infections with sensitive strains of bacteria. 293

Four studies evaluated the usefulness of different outcome measures for measuring quality of life after THR or revision THR, which showed that there was no consistency in the tools used to assess quality of life. Feeny et al. 294 reported that there is low agreement between certain outcome measures [SF-36, standard gamble, Health Utilities Index (HUI)-2 and HUI-3]. Dawson et al. 295 and Jones et al. 296 found that disease-specific measures reported larger changes than generic and utility measures. Ostendorf et al. 297 recommended the use of the OHS and the SF-12 in the assessment of THR and the EQ-5D in situations in which utility values are needed.

Overall, studies confirmed the long-standing claims that THR and RS are cost-effective interventions for patients with OA of the hip. However, there is little evidence from long-term trials on differences between implant brands and types of prostheses. This limits the conclusions that can be drawn with regard to the most cost-effective type of prosthesis. Studies used different methodologies to estimate costs (reference costs vs. prices actually paid by health-care centres) and definitions of costs included varied extensively, and many studies did not clearly report how costs were broken down. Although this review concentrates on clinical outcomes measured by the EQ-5D, the included studies tended to use more than one outcome measure with great variation across studies. In summary, THR, more so than RS, is a widely researched topic and receives great interest in many countries; however, further research should set out to include an assessment of the cost-effectiveness of different treatments.

Core studies for the cost-effectiveness analysis

Ranking eligible cost studies by year and country (most recent UK studies on top) and utility studies by number of participants, 11 studies were identified that were potentially useful to inform the decision model. These included one HTA and a further four cost-effectiveness studies. The HTA assessed the cost-effectiveness of hip RS compared with watchful waiting and THR. 19 The cost-effectiveness studies included three models that compared the cost-effectiveness of RS and THR,253 the cost-effectiveness of cemented, cementless and hybrid prostheses44 and the cost-effectiveness of two particular prosthesis types,38 respectively. One cost-effectiveness study was included that evaluated THR and RS but did not use a model. 40

The remaining six studies included partial economic evaluations that examined either costs or consequences but not both. Vanhegan et al. 292 reported costs for revision THR; Baker et al. 252 and Hulleberg et al. 258 reported medium- to long-term utilities in small populations; Dawson et al. 295 investigated quality of life post revision THR; and Bozic et al. 284 measured health state utilities for chronic OA of the hip, successful primary THR, failed primary THR, successful revision THR, failed revision THR and chronically infected THR. Rolfson et al. 298 evaluated the Swedish patient-reported outcomes data, reporting utilities for close to 35,000 THR patients.

Of the 11 studies three reported costs for THR,19,40,44 two reported costs for follow-up of successful THR19,40 and three reported costs of revision THR19,44,292 (see Appendix 12). Costs for RS were reported in three studies. 19,40,253 Of these, Edlin et al. 40 and Vale et al. 19 also reported follow-up costs after successful RS and Bozic et al. 253 reported costs for revision RS (see Appendix 11).

The studies reporting the most useful data on utilities following THR were those by Pennington et al. ,44 Rolfson et al. ,298 Hulleberg et al. ,258 Dawson et al. 295 and Bozic et al. 284 (see Appendix 14). Utilities for RS were reported in only three studies40,252,253 (see Appendix 15). No data were identified on quality of life at > 12 months post RS or for post-revision RS. Follow-up costs reported by Vale et al. 19 were the same for THR, RS and revision THR. Similarly, Bozic et al. 253 made no distinction between revision following THR or RS in terms of costs.

Quality assessment of core studies

Of the 11 core studies, five252,253,258,295,298 provided useful information on EQ-5D utility scores only and one292 provided useful data on costs only. These partial economic evaluations were not included in the critical appraisal. 309

Five studies19,38,40,44,253 were full economic evaluations and have been critically appraised using the CHEC-list. 249 Of these five studies, four19,38,44,253 included models. These studies have also been critically appraised using an adapted checklist for models developed by Philips et al. 250

Table 46 shows that all studies met ≥ 16 of the 19 criteria in the CHEC-list. 249

Table 47 shows that all studies met ≥ 20 of the 32 criteria for economic models provided by Philips et al. 250 All studies correctly reported the time horizon and the perspective of the model, and the inputs used within the models were consistent with the perspectives that were chosen. In terms of costs and outcomes used in the model, these were appropriate to the specific study data set that was used. All studies conducted subgroup analyses. None of the studies applied a half-cycle correction and no justification was given for its exclusion. In addition, Pennington et al. 44 did not provide a clear definition of all of the options under evaluation and Briggs et al. 38 did not specify the cycle length of the model.

Core studies for the economic model

Of the 11 core studies, Edlin et al. ,40 Pennington et al. ,44 Vale et al. 19 and Vanhegan et al. 292 provided data for the model in Chapter 9 (see Chapter 9 for the rationale of the selection procedure). This section will provide a brief description of the four core studies (Table 48).

Edlin et al. 40 reported a cost–utility analysis of RS compared with THR as part of a RCT of 126 adult patients with severe arthritis of the hip. Patients were randomised on a 1 : 1 basis between THR and RS. All RS patients received a Cormet™ (Corin Group, Cirencester, UK) metal-on-metal RS prosthesis. The THR patients received one of three types of prosthesis (ceramic-on-ceramic, metal-on-metal or metal-on-polyethylene) depending on the surgeon’s preference. The study took the NHS perspective and considered the within-trial period without any extrapolation past the 12-month trial period. The costs were reported in 2009/10 UK pounds and EQ-5D 3 Levels (EQ-5D-3L) outcomes were measured as secondary outcomes of the trial.

The study used Healthcare Resource Group v4 (HRG4) reference costs combined with NHS trust finance department list prices for implants and IPD on length of stay (LOS). Resource use data and personal costs were obtained from patient-reported data. Univariate sensitivity analyses included an assessment of the impact of using the cheapest THR type (metal-on-metal) for all THR operations. The study reported NHS and Personal Social Services (PSS) costs after 12 months by type of hip replacement (THR vs. RS), including the costs of the initial operation/care, subsequent inpatient, outpatient, primary and community care, aids and medication [THR £7217 (£1320); RS £6653 (£917)], as well as private and social costs. The main results of this analysis included a difference in QALYs of 0.033 in favour of RS after 12 months and a greater cost of RS (difference of £564) in the first 12 months following surgery. This resulted in an ICER for RS of £17,451 per QALY. These results are based on a short-term trial using a single RS prosthesis type. The study did not explore variation in costs within for each type of prosthesis used in THR. Variation in prosthesis costs by hospital, a change in current practice regarding the choice of THR implant, longer follow-up (including higher revision rates for RS than for THR) and use of different RS implants may affect the reported cost-effectiveness in this study.

Pennington et al. 44 used IPD from three data sources (national PROMs programme, the NJR and Hospital Episode Statistics) to compare the cost effectiveness of cemented, cementless and hybrid THR in adult patients with hip OA. They conducted a probabilistic Markov model over patients’ lifetime taking the NHS perspective. Implant prices were based on prices paid by English NHS centres. Costs for surgery plus hospital stay were taken from the literature and adjusted for LOS by prosthesis type and costs of revision were varied by reason for revision. Costs were reported as 2010/11 prices. The national data sources provided data on quality of life, LOS, rates of revision and rerevision and mortality for 30,203 patients.

Patients receiving different prosthesis types were matched by age, sex, number of comorbidities, ASA grade, BMI, deprivation, preoperative quality of life, surgeon experience and hospital type. The study reported data on the combined cost of the prosthesis, operating theatre and hospital stay, quality of life at 6 months post surgery and 5- and 10-year revision rates by prosthesis type, age group and sex. Overall, the study concluded that in patients aged 70 years the ICER for a hybrid prosthesis compared with a cemented prosthesis was £2100 for men and £2500 for women, with hybrid prostheses resulting in higher quality of life in all subgroups except women aged 80 years and cemented prostheses being the least costly option. The initial costs of a cementless prosthesis were highest in all subgroups. One of the limitations of the study was that it assumed that the observed quality of life at 6 months post surgery would remain unchanged for the patients’ lifetime. Furthermore, the study did not consider different revision rates by brand for the three different THR types.

Vale et al. 19 undertook an assessment of the clinical effectiveness and cost-effectiveness of RS compared with watchful waiting (i.e. patient monitoring, drug-based treatment and supportive activities including physiotherapy), THR and other bone-conserving treatments. The HTA comprised a systematic review of the clinical effectiveness and cost-effectiveness of RS compared with any of the treatments above and a Markov model comparing the comparators from the NHS perspective for patients suitable for RS for up to 20 years. Cost data (in 2000/1 UK pounds) for THR and revision THR were taken from the literature (£4195 and £6027, respectively) and prostheses costs for RS were obtained from manufacturers. The model considered the lower of the two RS implant costs obtained (£1730 vs. £1890), resulting in an overall cost of £5515 for RS. LOS was estimated to be 10 or 12 days for THR and 8 or 10 days for RS. All other costs including use of the operating theatre and staff, radiography, outpatient visits and first-year follow-up costs were assumed to be the same for RS and THR. First-year follow-up included two outpatient visits with one radiography scan, totalling £118.74. Quality-of-life estimates considered pain levels and quality-of-life scores for mild, moderate and severe OA and were combined with revision and mortality rates to generate QALYs.

The main conclusion from the systematic review was that evidence from the literature on the effectiveness of RS was limited. Revision rates were reported to range between 0% and 14% over a 3-year follow-up period for RS compared with ≤ 10% over 10 years for THR. Patients with RS experienced less pain than patients managed by watchful waiting. Results from the model showed that RS was dominated by THR based on assumptions about revision rates for RS and the lower cost of THR. In subsequent sensitivity analyses the revision rates for RS had to be reduced to < 80–88% of the THR revision rates before RS was no longer dominated by THR. However, RS dominated watchful waiting within the 20-year follow-up. The study was limited because of the lack of data for the parameters of the model, particularly revision rates for different RS brands and effectiveness data for revision THR following RS. Furthermore, available data for RS originated from a small number of surgeons.

Vanhegan et al. 292 investigated the costs of 305 consecutive revision THRs by reason for revision in 286 patients, with a diagnosis of hip OA in 64% of revisions (n = 195). Revision THR was carried out in a single tertiary centre by one of three experienced surgeons. Costs were obtained from the finance department of the tertiary centre (in 2007/8 UK pounds) and included costs of the implant, materials and augmentation, use of the operating theatre and recovery room, the inpatient stay and laboratory tests, radiology, pharmacy, physiotherapy and occupational therapy. The study provided cost data on 13 different implants and data on resource use and costs by reason for revision (aseptic loosening, deep infection, periprosthetic fracture and dislocation).

The mean costs of revision for aseptic loosening, deep infection, periprosthetic fracture and dislocation were reported to be £11,897 (SD £4629), £21,937 (SD £10,965), £18,185 (SD £9124) and £10,893 (SD £5476), respectively. Higher complication rates as well as reoperation rates were associated with revisions for deep infection, periprosthetic fracture and dislocation. However, the numbers of revisions for these three indications were relatively small (n = 76, n = 24 and n = 11, respectively). Although the cost estimates can be assumed to be very accurate, they are limited by their lack of generalisability as they were based on one single tertiary centre. Furthermore, the study did not consider the cost of readmission for complications and other direct and indirect medical and social costs.

Inclusion criteria

Exclusion criteria

• Abstract/conference proceedings, letters and commentaries.

• Non-English-language publications.

• < 1000 patients included in the registry study at the time of publication.

• Hip/knee data not reported separately.

Resurfacing arthroplasty

Eight registry studies provided evidence on RS. 15,311,313,318,329,349,351,353 The majority of these studies investigated various comparisons between THR and RS. Table 49 provides a summary of the RS studies.

Total hip replacement

In total, 22 registry studies reported evidence on THR, with the majority of these studies investigating various types of THR surgery or demographic differences regarding the specific countries. Table 50 provides a summary of the THR studies.

Core articles included in the economic model and survival analysis

The prioritisation of the eligible studies resulted in the identification of 30 papers that were deemed to be potentially useful for the economic model and survival analysis. The final number of core papers that helped to inform the survival analysis in this report was three. 15,16,318 This was in addition to the annual reports from the Swedish Arthoplasty Registry,96 the NJR36 and the AOANJRR,95 which were used for comparison of survival analysis methods.