MJA
MJA

Can we reduce disease burden from osteoarthritis?

Leonie Segal, Susan E Day, Adam B Chapman and Richard H Osborne
Med J Aust 2004; 180 (5): S11. || doi: 10.5694/j.1326-5377.2004.tb05907.x
Published online: 1 March 2004

Abstract

  • The comparison of disparate interventions for the prevention and management of osteoarthritis (OA) is limited by the quality and quantity of published efficacy studies and the use of disparate measures for reporting clinical trial outcomes.

  • The “transfer to utility” technique was used to translate published trial outcomes into a health-related quality-of-life (utility) scale, creating a common metric which supported comparisons between disparate interventions.

  • Total hip replacement (THR) and total knee replacement (TKR) surgery were the most effective treatments and also highly cost-effective, at estimated cost per quality-adjusted life-year (QALY) of $7500 for THR and $10 000 for TKR (best estimate).

  • Other apparently highly cost-effective interventions were exercise and strength training for knee OA (< $5000/QALY), knee bracing, and use of capsaicin or glucosamine sulfate (< $10 000/QALY).

  • The cost per QALY estimates of non-specific and COX-2 inhibitor non-steroidal anti-inflammatory drugs were affected by treatment-related deaths and highly sensitive to the discounting of life-years lost.

  • OA interventions that have been shown to be ineffective (eg, arthroscopy) are targets for redistribution of healthcare resources.

  • OA interventions which lack efficacy studies (eg, prevention programs) require further research to assist priority setting.

  • The application of the Health-sector Wide model to OA demonstrates its role as an evidence-based model that can be successfully applied to identify marginal interventions — those to be expanded and contracted to reduce the expected burden of disease, within current healthcare resources.

Osteoarthritis (OA) affected over 1.2 million Australians in 19951 and is associated with both substantial loss of quality of life and high management costs. In 2001, these costs were estimated at $1090 million.2

There is a wide range of options for managing OA. These include prescription drugs, over-the-counter medications, exercise and strength training, surgery, patient education and complementary therapies. Recognising that distortions in funding arrangements for the delivery of healthcare result in highly disparate cost-effectiveness ratios,3-5 logic suggests there are opportunities for health gain in redistributing resources from services that perform poorly (high cost per unit of benefit) to those that perform better (low cost per unit of benefit). Priority-setting models are designed to identify such desirable resource shifts, and one model that has been found to be appropriate for national or regional decision-making is the Health-sector-Wide (HsW) disease-based model.6,7 This model, like the Quality Adjusted Life Year (QALY) league table approach and Program Budget Marginal Analysis, involves comparing marginal cost-effectiveness ratios of possible interventions.7 The model also insists on objective evidence for determining effectiveness, a defined process for selecting interventions to ensure comprehensiveness, and a means for staging the analysis through diseases and disease stages. The model was recently applied to investigating how the disease burden of OA might be reduced by adjusting the health service mix.2

Methods

Applying the model involved three broad research tasks:

  • identifying all potential interventions (service types) for preventing and managing OA and, through a literature review, selecting a subset for inclusion in the model;

  • conducting economic evaluations of the selected interventions, using published evidence on outcomes and costs; and

  • comparing performance of the interventions (primarily in terms of cost-effectiveness) and developing policy recommendations on desirable resource shifts.

An advisory panel of clinicians, health department officers, consumers and consumer organisation representatives (see Acknowledgements) was established to support the research team, ensuring appropriate coverage of the literature and application of the model.

Comparison of outcomes from disparate studies

The variety of disease-specific and generic instruments that have been used to report the effectiveness of OA treatments makes it difficult to compare published studies. HsW modelling required a means to compare each intervention on the same scale. Ideally, the method for doing this should also enable global comparisons, and combine quality of life and mortality. Health-Related Quality of Life, based on “utility” and scaled between 0 (death) and 1 (best possible health), meets this requirement. Outcomes expressed as utility values can be combined with time lived in particular health states and added to mortality rate differentials to calculate effectiveness in quality-adjusted life-years (QALYs). In this way, the performance of all interventions can be measured in the same metric (cost/QALY) and directly compared. Utility values can be obtained from a multi-attribute utility instrument such as the European EQ-5D30 or the Australian Assessment of Quality of Life (AQoL) instrument.31,32 Conversion of the SF-3633 to a utility scale has been the subject of ongoing research, but the initial algorithm appeared to truncate at the severe end of the scale and was therefore unsuitable for our purposes.34,35 While this conversion has since been revised, it requires access to the full SF-36 dataset, which is an impediment to its use.36 The use of disability weights to calculate disability-adjusted life-years (DALYs) was also rejected, as the published disability weights for OA describe only three levels, which did not map onto trial data.37

For this study, commonly used OA outcome scales were converted into a utility-equivalent scale. The technique, termed “transfer to utility”, or TTU, involved administering the AQoL alongside common OA outcome instruments such as the SF-36, a visual analogue scale (VAS) for pain, and the Western Ontario and McMaster Arthritis Index38 in 303 people with OA. Participants were recruited from rheumatology clinics, orthopaedic waiting lists and the Arthritis Foundation of Victoria to ensure a wide range of severity of OA was captured. The survey was approved by the Royal Melbourne Hospital Human Ethics Committee.

Equivalent utility values (based on the AQoL) for the selected outcome instruments were estimated from these data by means of multiple regression analysis. These relationships were applied to outcomes reported in published studies for both the intervention and control cohorts to estimate utility gain attributable to interventions.

The technique can be illustrated by applying it to total hip replacement (THR), where the reported outcomes are the SF-36 subscale scores, measured before and after surgery. A multiple linear regression model of the AQoL (dependent variable) was constructed as a function of the eight subscales of the SF-36. The relationship between the observed and predicted scores is shown in Box 2, and shows the good explanatory power of the model (adjusted R2 = 0.66; P < 0.001). Separate weights were obtained for hip OA (R2 = 0.70; P < 0.001) and knee OA (R2 = 0.63; P < 0.001). Box 3 shows hip-specific parameters and the estimation of utility-equivalent outcomes from the chosen THR study,28 showing a utility gain 12 months after surgery of 0.304. Further details of the “transfer to utility” technique are described elsewhere.2

Downstream impacts/discounting

Future impacts (costs and benefits) were calculated where relevant and when evidence was available. For THR and total knee replacement (TKR) surgery, benefits and costs (including costs for revision surgery) were modelled over 15 years, knee bracing was modelled over 18 to 36 months and exercise over 12 months. Primary care was modelled over 20 years, with estimated mortality based on all-cause mortality, and quality-of-life gain based on the reduced incidence of OA.

Because there has been debate about the desirability of discounting health,39 we derived two sets of cost-effectiveness estimates, using discount rates on health benefits of 5% per annum and zero, but costs always discounted at 5% per annum. In the United Kingdom, future life-years are discounted at 1.5%–2.0% per annum, and future costs at 6% per annum (the status until 2003, with the revised standard discounting of costs and benefits at 3.5% per annum),40,41 while the Australian Pharmaceutical Benefits Advisory Committee guidelines for listing of drugs on the Pharmaceutical Benefits Scheme (PBS) suggest costs and benefits be discounted at 5% per annum.42

Results

Results are reported as utility benefit, cost per individual and cost/QALY in Box 5.

Surgery: The most effective intervention modelled was THR. The estimated utility gain from THR was 0.305. This indicates a substantial improvement in quality of life, considerably greater than that associated with other interventions modelled. As benefits persist for several years (except for the 12%–20% of patients expected to receive revision surgery),43,44 estimated QALY gain over a lifetime from THR was 3.52 per hip (adjusting for expected surgical mortality of 1/1000 and all-cause mortality in the cohort), and 2.086 per knee for TKR. At a weighted cost of primary and revision surgery of $16 000–$17 000 per case, cost/QALY for THR was $5000 (future benefits undiscounted) or $7000 (discounted at 5%), and, for TKR, a cost/QALY of $8000 (future benefits undiscounted) or $11 000 (discounted at 5%). Cost/QALY was not calculated for arthroscopy in knee OA, as recent randomised controlled trials (RCTs) suggest that the treatment is equivocal.29,45

Non-steroidal anti-inflammatory drugs (NSAIDs): Non-specific NSAIDs and COX-2 NSAIDs were found to perform similarly in terms of outcomes and side effects, based on seminal trials and a report by the United States Federal Drugs Administration which interpreted side-effect data.46 NSAIDs modelled were diclofenac, naproxen and celecoxib, using doses reported in the key trials. Based on three placebo/paracetamol-controlled trials,22-24 estimated mean utility benefit compared with placebo was 0.043. This benefit was offset by excess mortality equal to 0.044 or 0.029 QALYs (future life-years undiscounted or discounted at 5% per annum, respectively). Net QALY gain was close to zero or negative if future life-years from current deaths were undiscounted and 0.014 if discounted at 5% per annum, resulting in a wide estimate of cost/QALY from $15 000 to “infinity” for non-specific NSAIDs and between $33 000 and infinity for COX-2-specific NSAIDs. In most scenarios, COX-2 NSAIDs are dominated by non-specific NSAIDs.

Primary prevention: Because of the absence of RCTs for the primary prevention of OA through weight loss (or other means), this model used epidemiological data on the prevalence of OA and obesity, incorporating evidence of the effectiveness of weight loss interventions.1,8-10,12 Cost/QALY estimates ranged from $2000 to $48 000, depending on the type of program. But, in the absence of intervention trials, this analysis is speculative and requires testing through intervention trials. Primary prevention through modifying other risk factors, such as recreational or work-related injury, could not be established owing to lack of evidence.

Patient education: Although community-based education programs are widely applied, there is little published quantitative evidence. Group programs, including those run by lay leaders13,14 and health professionals,15,16 were reviewed, but cost/QALY estimates could not be generated as inconsistent findings made the evidence for effectiveness difficult to interpret.

Knee brace: Specifically manufactured and fitted knee braces for people with knee OA were found to be effective and cost-effective. Cost/QALY was $4000–$12 000, depending on the length of time the brace was worn.17

Exercise and strength training: Intensive exercise and strength training based in primary care20 or outpatient clinics21 was highly cost-effective ($3000–$15 000/QALY) and somewhat less cost-effective for an intensive home-based program ($10 000–$34 000/QALY).19 The evidence for the effectiveness of a less intensive home-based program was equivocal.18

Other pharmacotherapies: Topical capsaicin (for knee, hand, elbow and ankle OA) and glucosamine sulfate were found to be cost-effective at less than $5000/QALY. These drugs had equivalent efficacy to NSAIDs, with no evidence of excess morbidity or mortality.25,26,47 The performance of avocado/soy unsaponifiables was unclear because of conflicting results from the small number of published reports.48 Results for these therapies should be considered provisional, as they are based on only a few RCTs. However, recent trials lend support to a role for glucosamine sulfate.49,50

Discussion

An important contribution of this priority-setting exercise is the development of an empirically driven method for comparing interventions, based on the published clinical trial literature, even where different outcome instruments are used. The “transfer to utility”, or TTU, allowed disparate trial outcome data to be translated into a common metric — a utility equivalent score — and the incorporation of mortality and costs to ascertain performance in terms of cost/QALY. This technique placed all treatments on a “level playing field”. The HsW model incorporating this technique should be less prone to the biases of more subjective priority-setting techniques, such as Program Budget Marginal Analysis, which allow “expert opinion” a greater role.6,7 The inclusion of the Advisory Panel in the process was found to be valuable for identification and interpretation of the literature, but not in development of the evidence base.

The quality of the advice resulting from the HsW model is dependent on the quality of clinical evidence. The modelling highlighted, in a systematic way, critical data gaps and future research priorities. Given the gaps in the data, outcomes and costs had to be modelled. This involved adopting assumed values and some limitation in scope, the latter reflecting the breadth of a priority-setting exercise relative to the research resources available. Most important of these limitations was the failure to conduct analyses at the subpopulation level, effectively assuming that mean incremental cost–utility ratio was a reasonable approximation across the relevant clinical population. Further, the scope of effects has been restricted to health benefits (defined by QALYs) and healthcare costs related to individuals with OA, a common (although not universal) assumption of cost-effectiveness analyses. In terms of technical simplifications, costs and benefits for a cohort should be based on the actual age distribution, rather than assuming a cohort of mean age consistent with the clinical trial information as we did in this study. Further, multivariate probabilistic sensitivity analysis is preferable to the use of univariate sensitivity analysis.

Despite these limitations, a number of conclusions can be developed from the HsW analysis, especially where clinical trial evidence is strong and the range of cost-utility estimates is narrow across plausible parameter values (Box 6). In drawing conclusions from the model it is presumed that cost-effectiveness is the sole or primary decision criterion. Given this premise, and the strong evidence of effectiveness and favourable cost–utility ratios for THR, TKR, intensive clinic-based exercise and strength training, and knee bracing, providing these services to all those for whom the treatment is clinically appropriate would represent an efficient use of healthcare resources. Subject to confirmatory evidence for efficacy and side-effect profile, topical capsaicin and oral glucosamine may also fit this category, with more modest effectiveness compensated by lower costs. The need for better targeting of COX-2 NSAIDs to subpopulations with greater capacity for net benefit is suggested by the wide range of cost–utility estimates (relative to placebo), and dominance by non-specific NSAIDs in many scenarios. Our analysis suggests that the resources released could be redirected to interventions that yield greater benefit per unit cost to reduce the overall disease burden of OA. Additional research into NSAIDs to highlight the performance of specific formulations and the impact on identifiable subpopulations is indicated. Independent of the HsW model, the literature suggests that the wide application of some treatments, such as arthroscopy for knee OA, may be a poor use of resources.29,45

It can be seen that economic analyses such as this can provide structured support to policy, and can complement clinicians’ judgement. How to achieve any agreed resource shifts and the optimal size of redistributions is properly the subject of a separate study. The translation of priority-setting recommendations into policy, and ultimately into health improvements for the population, will require the involvement of policymakers and clinical communities. The Bone and Joint Decade and its commitment to reducing the burden of disease from OA may contribute to this process.

1: Interventions for prevention and management of osteoarthritis (OA) that were included in the modelling

Primary prevention

  • Weight loss

Comprehensive media/community campaign;8,9 intensive primary care (GP and nurse) diet and behavioural weight loss intervention for a general overweight group;10 intensive primary care weight loss intervention for people with previous knee injury;10,11 surgery for obesity (eg, gastric banding)12


Patient management

  • Patient education

Lay-led program;13,14 GP/clinical nurse educator-led program15,16

  • Physical therapies

Specially fitted knee brace for person with OA of the knee17

  • Exercise/strength training

Home-based basic program;18 home-based intensive program;19 primary care clinic-based intensive program;20 out-patient-based intensive program21

  • Pharmacotherapies

    • Prescription/over-the-counter medications

      Non-specific non-steroidal anti-inflammatory drugs (NSAIDs) (diclofenac and naproxen);22,23 COX-2-specific NSAIDs (celecoxib)22,24

    • Complementary medicines

Oral glucosamine sulfate;25 ASU (avocado/soybean unsaponifiables);26 topical capsaicin27

  • Surgery

Total knee replacement;28 total hip replacement;28 knee arthroscopy with lavage29

3: Conversion of SF-36 subscale scores to a utility score for osteoarthritis of the hip

Before surgery


12 months after surgery


SF-36 subscale

Regression weight

SF-36 score*

Adjustment

Utility score

SF-36 score*

Adjustment

Utility score


(Constant)

– 0.1976

Physical function

  0.4803

26.9

0.4661

  0.224

66.6 

0.7345

  0.353

Role physical

– 0.0001

14.6

– 0.002

58.7

– 0.007

Bodily pain

  0.2438

32.9

0.4562

  0.111

72.8

0.7486

  0.182

General health

  0.0003

66.3

  0.023

73.8

  0.025

Vitality

  0.0009

47.1

  0.045

67.5

  0.064

Social function

  0.0018

52.5

  0.092

88.6

  0.155

Role emotion

– 0.0003

60.0

– 0.018

71.8

– 0.022

Mental health

  0.0026

71.4

  0.187

81.7

  0.214

Estimated utility value§

  0.464

  0.767

Utility change

  0.304


* Source: Bachmeier, March and Cross.28 † The physical function and bodily pain scales undergo an initial non-linear transformation. The physical function score of 26.9 is transformed by means of a quadratic adjustment: 0.108164 + (0.01595 × 26.9) – (0.000098 × 26.92) = 0.4663, which is multiplied by the regression weight of 0.4803 to obtain the utility component of 0.224. ‡ SF-36 subscale is multiplied by the subscale-specific regression weight. For example, the mental health score of 71.4 is multiplied by the regression weight of 0.0026 to obtain a “utility component” score of 0.187. § The utility component of each subscale is summed with the constant to obtain a total “utility equivalent” score.

4: Key assumptions underlying the modelling

Attribute and modality

Key assumption


Duration of benefits

Primary prevention

20 years

Education/self-management

2 years

Exercise

1 year

Knee brace

18 months to 3 years

Total hip and knee replacement

15 years, adjusted for all-cause mortality

  • Total hip replacement cohort: surgery at 70 (women) and 66 (men) years of age

  • Total knee replacement cohort: surgery at 70 (women) and 71 (men) years of age

Revision surgery, 12% or 20%, spread evenly over 15 years

Non-steroidal anti-inflammatory drugs*

12 months (treatment contiguous with benefit)

Natural therapies

12 months (treatment contiguous with benefit)

Discount rate all modalities

Downstream costs/cost savings

Beyond 12 months, discounted at 5% per annum

QALY gains or losses

0 (undiscounted) 5% per annum

Excess morbidity

Non-steroidal anti-inflammatory drugs*

Hospitalisation:

  • Congestive heart failure, 0.5/100 patient-years of NSAID use

  • PUBs, 2.13/100 patient-years of NSAID use

Excess mortality

Non-steroidal anti-inflammatory drugs*

Death rate:

  • The same across all NSAIDs*

  • Congestive heart failure, 5%; morbidity, 0.00025 per patient-year of NSAID use

  • Gastrointestinal mortality, 10%; hospitalisation for PUBs, 0.00213 per patient-year of NSAID use

Translation of deaths to life-years: mortality × 18.5 life-years (mean life expectancy, age 65 years)

Total hip and knee replacement

Death rate from surgery, 1/1000 

Sensitivity analyses

Discount rate of QALYs, 0 and 5% Clinical trial results: ± 1 SD (NSAIDs) or ± 2 SDs (other)


NSAID = non-steroidal anti-inflammatory drug. QALY = quality-adjusted life-year. PUBs = perforations, peptic ulcerations, gastric bleeding. * Diclofenac, naproxen, celecoxib.

5: Key results of cost–utility analyses

Program

Mean QALY gain per person* (zero discount rate)

Mean program cost per person

Cost/QALYbest estimate


Primary prevention

1. Comprehensive mass media program for weight loss

$4

$20 000

2. Intensive primary care weight loss program delivered by GP or dietitian for overweight or obese people

0.09 

$720

$11 000

3. Intensive primary care weight loss program delivered by GP or dietitian for overweight or obese people with previous knee injury

0.08

$720

$12 000

4. Surgery for obese people

1.05

$15 000

$20 000

Management

Education

5. Lay-led group education

0.072

$162

$2 400 to ∞

6. Primary care: GP or clinical nurse educator plus phone support

Equivocal

$200–$400

Exercise/strength training

7. Home-based exercise — basic

0.022

$400

$18 000 to ∞

8. Home-based exercise — intensive

0.100

$1 420

$15 000

9. Clinic-based exercise — primary care

0.091

$480

$5 000

10. Clinic-based exercise — outpatients

0.078

$590

$8 000

Knee brace

11. Specially fitted knee brace

0.12 to − 0.355

$1 300

$6 000

Pharmacotherapies: prescription or over-the-counter medications

12. Non-specific NSAIDs (naproxen, diclofenac)

+ 0.043 – 0.029 to 0.044§

Drug, $104/year; Morbidity, $70

$15 000 or ∞

13. COX-2-specific NSAIDs (celecoxib)

+ 0.043– 0.029 to 0.044§

Drug, $391/year; Morbidity, $70 

$33 000 or ∞

Complementary medicines

14. Glucosamine sulfate

0.052 

$180

$3 000

15. Avocado/soy unsaponifiables

0.081

$333

$5 000 to ∞

16. Topical capsaicin

0.053

$236

$5 000

Surgery

17. Total knee replacement

+ 2.086 – 0.015§

$16 500

$8 000 or $11 000 

18. Total hip replacement

+ 3.52 – 0.015§

$16 500 

$5 000 or $7 000

19. Knee arthroscopy with lavage

Equivocal

$3 500


* Relative to placebo, or to no intervention, for total hip and knee replacement surgery. † Rounded to nearest $1000; both costs and benefits discounted at 5% unless otherwise stated. ‡ Benefit from symptom control. § Mortality from cardiac and gastrointestinal side effects for NSAIDs or perioperative mortality for total hip replacement, total knee replacement, deducted from QALY gain. ¶ Zero discounting of future life-years lost from current deaths. QALY = quality-adjusted life-year. NSAIDs = non-steroidal anti-inflammatory drugs. ∞ = No evidence of benefit, possible negative benefit.

6: Resource implications for the prevention and management of osteoarthritis based on cost-utility analyses

Highly cost-effective programs (cost/QALY < $15 000) supported by strong evidence

  • Intensive clinic outpatient-based exercise/strength training programs

  • Total hip replacement surgery

  • Total knee replacement surgery

Policy implication: Programs/service level should meet all clinically indicated need.

Programs that appear highly cost-effective (cost/QALY < $10 000), but are based on limited evidence

  • Topical capsaicin

  • Glucosamine sulfate

  • Knee brace

Policy implication: Programs/services might warrant expansion to match clinical need. A program to collate/gather additional trial evidence should be supported. In the interim, greater access in response to clinical need appears appropriate.

Interventions which may be highly cost-effective (cost/QALY < $20 000), but based on indirect evidence

  • Primary prevention through weight loss in obese persons through comprehensive community-based media campaign, intensive primary care intervention, surgery for persons who are seriously obese (to determine impact on incidence of OA)

  • Policy implication: A research program should be supported, possibly through pilot implementation and/or well constructed observational studies.

Interventions for which cost-effectiveness is uncertain and highly sensitive to assumptions of the model (and the specific formulation)

  • Non-specific non-steroidal anti-inflammatory drugs

  • COX-2-specific non-steroidal anti-inflammatory drugs

  • Patient education — lay-led models

Policy implication: Explore assumptions of the model and cost-effectiveness in specific patient groups (and with particular formulations); redirect resources to patient groups where higher net benefit can be realised.

Interventions where efficacy is equivocal or not demonstrated

  • Patient education — professionally led models

  • Arthroscopy to manage knee OA

Policy implication: Services to be curtailed and resources redirected to areas with demonstrated effectiveness.

QALY = quality-adjusted life-year. OA = osteoarthritis.

Online responses are no longer available. Please refer to our instructions for authors page for more information.