Abstract

Objectives. Bell's palsy (BP), which causes facial paralysis, affects 11–40 people per 100 000 per annum in the UK. Its cause is unknown but as many as 30% of patients have continuing facial disfigurement, psychological difficulties and occasionally facial pain. We present an randomised controlled trial (RCT)-based economic evaluation of the early administration of steroids (prednisolone) and/or antivirals (acyclovir) compared to placebo, for treatment of BP.

Methods. The RCT was not powered to detect differences in the cost-effectiveness; therefore, we adopted a decision analytic model approach as a way of gaining precision in our cost-effectiveness comparisons [e.g. prednisolone only (PO) versus acyclovir only versus prednisolone and acyclovir versus placebo, prednisolone versus no prednisolone (NP) and acyclovir versus no acyclovir]. We assumed that trial interventions affect the probability of being cured/not cured but their consequences are independent of the initial therapy. We used the percentage of individuals with a complete recovery (based on House–Brackmann grade = 1) at 9 months and Quality Adjusted Life Years (e.g. derived on responses to the Health Utilities Index III) as measures of effectiveness. Other parameter estimates were obtained from trial data.

Results. PO dominated—i.e. was less costly and more effective—all other therapy strategies in the four arms model [77% probability of cost-effective (CE)]. Moreover, Prednisolone dominated NP (77% probability of being CE at £30 000 threshold) while no acyclovir dominated aciclovir (85% chance of CE), in the two arms models, respectively.

Conclusions. Treatment of BP with prednisolone is likely to be considered CE while treatment with acyclovir is highly unlikely to be considered CE. Further data on costs and utilities would be useful to confirm findings.

Introduction

Bell's palsy (BP) is an acute unilateral paralysis of the facial nerve. 1 Its cause is unknown but it affects 11–40 people per 100 000 in the population per annum, most commonly in the age group 30––45. 2 The condition is most common among pregnant women and people who have diabetes, influenza, a cold or some other upper respiratory ailment. Although most recover, as many as 30% of people have a poor recovery with continuing facial disfigurement, psychological difficulties and sometimes facial pain. 2–4 In the absence of an established aetiology, treatment continues to be based upon the established pathophysiology: swelling and entrapment of the nerve.

Two Cochrane reviews have examined the effectiveness of oral prednisolone and aciclovir for the treatment of BP 5 , 6 and both report insufficient evidence on the effectiveness. In addition, high-dose steroid therapy has numerous potential side effects including peptic ulceration, hypertension and confusional states. 7 Antiviral therapy is expensive and it has been argued that it should be reserved for circumstances where definite benefits are likely to be obtained.

Given this lack of evidence, the UK NHS R&D Health Technology Assessment Programme commissioned an randomised controlled trial (RCT) to determine whether prednisolone or aciclovir, used separately or in combination and used early in the course of BP, is an effective and efficient treatment. The aim of this paper is to report evidence on the relative efficiency of these therapies.

Methods

Details of the RCT have been reported elsewhere. 8 Briefly, this was a multi-centre, double-blind, placebo-controlled, randomized, factorial trial involving patients with BP who were recruited within 72 hours after onset of symptoms. Five hundred and fifty-one patients were recruited from primary care settings and referred to 17 hospitals in Scotland between June 2004 and June 2006, where eligible patients were randomly assigned to receive 10 days of treatment of the following: 25 mg twice daily with prednisolone ( n  = 138) or 400 mg five times daily with acyclovir ( n  = 138), both agents ( n  = 134) or placebo ( n  = 141). Follow-up was 9 months. The primary outcome was complete recovery of facial function as rated on the House–Brackmann scale. Secondary outcomes included quality of life, appearance, pain, costs and relative efficiency. The study included adults of 16 years or older with unilateral facial weakness of no identifiable cause who presented to primary care or emergency department and could be referred to a collaborating otorhinolaryngologist within 72 hours of the onset of symptoms. Exclusion criteria were pregnancy, breastfeeding, diabetes, peptic ulcer disease, suppurative otitis media, herpes zoster, multiple sclerosis, systemic infection, sarcoidosis and other rare conditions and an inability to provide informed consent. All participants gave written informed consent. The study was approved by the Multicentre Research Ethics Committee for Scotland.

The economic evaluation conducted as part of this trial adopted a modelling approach as a means of gaining precision in cost-effectiveness estimates. Decision analytic models were constructed to compare the relative efficiency of the four randomized arms and also for the two randomized comparisons from the 2 × 2 factorial design: prednisolone against no prednisolone (NP) and aciclovir against no aciclovir. As the time horizon of the economic analyses mirrors that of the original study, we believe that this type of model gives a well representation of the decision problem. An example of the model structure is shown in Figure 1(a) . Within these models, it is assumed that the different trial interventions affect the probability of being cured or not cured and the consequences are assumed to be independent of the assigned therapy.

F IGURE 1

(a) Decision tree model for early treatment for BP: prednisolone alone versus aciclovir alone versus prednisolone + aciclovir versus no treatment (placebo). The four decision branches reflect the four groups provided by the 2 × 2 factorial trial design. It has been assumed that the costs consequences of being cured or not cured are independent of the initial treatment a person was allocated to. (b) BP decision tree model: prednisolone versus NP

F IGURE 1

(a) Decision tree model for early treatment for BP: prednisolone alone versus aciclovir alone versus prednisolone + aciclovir versus no treatment (placebo). The four decision branches reflect the four groups provided by the 2 × 2 factorial trial design. It has been assumed that the costs consequences of being cured or not cured are independent of the initial treatment a person was allocated to. (b) BP decision tree model: prednisolone versus NP

Parameter estimates used in the model

Parameter estimates for probabilities, costs and effectiveness required to populate the model were developed from trial data. These data related to risk of being cured or not cured at different time points, health services resource use and costs and health state utilities.

Probability of cure and not cure

Table 1 shows the proportion of subjects cured and not cured at 3 and 9 months, used as probabilities within the model. Normal probability distributions were attached to the difference in proportions between groups to allow for parameter uncertainty.

T ABLE 1

Probability parameters

 Probability being cured at 3 months Probability being cured at 9 months given not cured at 3 months 
Four arms model   
    Prednisolone alone (s.e.) 0.84 (0.03) 0.71 (0.11) 
    Aciclovir alone (s.e.) 0.60 (0.04) 0.44 (0.07) 
    Aciclovir and prednisolone (s.e.) 0.78 (0.04) 0.68 (0.09) 
    Placebo alone (s.e.) 0.65 (0.04) 0.57 (0.08) 
Prednisolone versus NP model   
    Prednisolone 0.83 0.49 
    NP 0.64 0.61 
    Difference (95% CIs) 0.19 (0.12 to 0.27)  
    PD assumed for difference Normal  
Aciclovir versus no aciclovir model   
    Aciclovir 0.71 0.49 
    No acyclovir 0.76 0.61 
    Difference (95% CIs) −0.05 (−0.12 to 0.03)  
    PD assumed for difference Normal  
 Probability being cured at 3 months Probability being cured at 9 months given not cured at 3 months 
Four arms model   
    Prednisolone alone (s.e.) 0.84 (0.03) 0.71 (0.11) 
    Aciclovir alone (s.e.) 0.60 (0.04) 0.44 (0.07) 
    Aciclovir and prednisolone (s.e.) 0.78 (0.04) 0.68 (0.09) 
    Placebo alone (s.e.) 0.65 (0.04) 0.57 (0.08) 
Prednisolone versus NP model   
    Prednisolone 0.83 0.49 
    NP 0.64 0.61 
    Difference (95% CIs) 0.19 (0.12 to 0.27)  
    PD assumed for difference Normal  
Aciclovir versus no aciclovir model   
    Aciclovir 0.71 0.49 
    No acyclovir 0.76 0.61 
    Difference (95% CIs) −0.05 (−0.12 to 0.03)  
    PD assumed for difference Normal  

PD, probability distribution.

Health care resource use and costs

The costs estimates used in the model were based on the cost of the initial treatments and follow-up costs. Follow-up costs included the use of resources in primary and secondary care and the subsequent use of other medications. These resources were costed using readily available unit costs.

Treatment costs

The doses and length of treatment for trial medications were defined by the trial protocol. The unit costs were obtained from British National Formulary (BNF) 9 ( Table 2 ).

T ABLE 2

Resource use and costs: drug treatment

Drug Dose Cost Note  BNF web page a 
Prednisolone 50 mg/day × 10 days 4.32 Prednisolone, tablets, 25 mg, 56-tab pack = £12.09 http://www.bnf.org/bnf/bnf/53/4259.htm 
Aciclovir 2000 mg/day × 10 days 6.57 Aciclovir, tablets, 400 mg, 56-tab pack = £7.31; 800 mg, 35-tab pack = £9.22 http://www.bnf.org/bnf/bnf/53/37356.htm 
Drug Dose Cost Note  BNF web page a 
Prednisolone 50 mg/day × 10 days 4.32 Prednisolone, tablets, 25 mg, 56-tab pack = £12.09 http://www.bnf.org/bnf/bnf/53/4259.htm 
Aciclovir 2000 mg/day × 10 days 6.57 Aciclovir, tablets, 400 mg, 56-tab pack = £7.31; 800 mg, 35-tab pack = £9.22 http://www.bnf.org/bnf/bnf/53/37356.htm 
a

Accessed on May 21, 2007.

Follow-up costs

Primary and secondary care resource use.

Health Care resources used were collected from primary care case notes in a convenience sample of 74 study subjects on any contacts made with health services or resources used by trial participants. This subsample appears representative of the whole study sample (see Supplementary Data available at Family Practice online for details of baseline characteristics). The number of contacts are described in Table 3 split by whether the data referred to cured or not cured patients (see Supplementary Data available at Family Practice online for further details on Resource Use). Hospital-based services (inpatient days, day cases and outpatient visits) were costed using data from the Information Services Department for Scotland 10 after deducting overheads allocated to the particular cost category (see Supplementary Data available at Family Practice online). Unit costs for primary care-based services were obtained from Curtis and Netten 11 and from the BNF for medications. 9

T ABLE 3

Resource use and costs: health care resource use by main cost categories, Cured or not cured patients

Concept  Cured at 3 months
 
Cured at 9 months
 
Not cured
 
Primary care (contacts) Hospital (inpatient days and day cases) Hospital outpatient (visits) Primary care (contacts) Hospital (inpatient days and day cases) Hospital outpatient (visits) Primary care (contacts) Hospital (inpatient days and day cases) Hospital outpatient (visits) 
N 52 53 51 11 11 10 
Mean (SD) 2.15 (3.9) 0.11 (0.38) 0.49 (1.17) 3.82 (3.68) 0.09 (0.3) 1.8 (2.62) 3.22 (2.73) 0 (0) 1.22 (1.56) 
Median (IQR) 1 (0–2) 0 (0–0) 0 (0–1) 3 (1–5) 0 (0–0) 0.5 (0–3) 3 (2–3) 0 (0–0) 1 (0–1) 
Concept  Cured at 3 months
 
Cured at 9 months
 
Not cured
 
Primary care (contacts) Hospital (inpatient days and day cases) Hospital outpatient (visits) Primary care (contacts) Hospital (inpatient days and day cases) Hospital outpatient (visits) Primary care (contacts) Hospital (inpatient days and day cases) Hospital outpatient (visits) 
N 52 53 51 11 11 10 
Mean (SD) 2.15 (3.9) 0.11 (0.38) 0.49 (1.17) 3.82 (3.68) 0.09 (0.3) 1.8 (2.62) 3.22 (2.73) 0 (0) 1.22 (1.56) 
Median (IQR) 1 (0–2) 0 (0–0) 0 (0–1) 3 (1–5) 0 (0–0) 0.5 (0–3) 3 (2–3) 0 (0–0) 1 (0–1) 

Total costs

Using all the data described above, estimates of the total mean costs for those cured and not cured were estimated ( Table 3 ). A simple ordinary least squared (OLS) regression was fitted to the data obtained from those people for whom data were able to be collected ( n  = 74). The total mean values used within the model were £210 (s.e. 58) for cured with not cured being £105 (s.e. 112) more costly at 3 months. Normal distributions were added to the total cost of being cured and not cured with the total cost of not cured bounded at zero within the probabilistic sensitivity analysis.

Estimation of utilities

The RCT 8 collected data on Health Utilities Index mark III (HUI III) 12 at baseline, 3 months and, if trial participants were not cured at 3 months, also at 9 months. Two analyses of covariance adjusting for baseline HUI III scores were used to obtain utility weights for participants who were cured and not cured at 3 and 9 months ( Table 4 ). In order to reflect the statistical imprecision surrounding these estimates when used in the model, normal distributions were attached to the mean difference in values based upon the results of a regression analysis.

T ABLE 4

HUI III regression analysis for three and nine months cured and not cured utility weights

Dependent variable: HUI III at three months 
Number of observations = 487 
 Coefficient Standard error 95% CI 
Constant 0.6146 0.0235 0.5684 0.6609 
Cured 0.0574 0.0132 0.0314 0.0834 
Dependent variable: HUI III at three months 
Number of observations = 487 
 Coefficient Standard error 95% CI 
Constant 0.6146 0.0235 0.5684 0.6609 
Cured 0.0574 0.0132 0.0314 0.0834 
Dependent variable: HUI III at nine months 
Number of observations = 137 
 Coefficient Standard error 95% CI 
Constant 0.5265 0.0495 0.4287 0.6243 
Cured –0.0019 0.0293 –0.0599 0.0561 
Dependent variable: HUI III at nine months 
Number of observations = 137 
 Coefficient Standard error 95% CI 
Constant 0.5265 0.0495 0.4287 0.6243 
Cured –0.0019 0.0293 –0.0599 0.0561 
Utility weights (mean values) 
Cured at 3 months Cured at 9 months Not cured 
0.9947 0.9900 0.9919 
Utility weights (mean values) 
Cured at 3 months Cured at 9 months Not cured 
0.9947 0.9900 0.9919 
Baseline characteristics. HUI III data All participants 
Mean: 0.786 sd 0.216 
Baseline characteristics. HUI III data All participants 
Mean: 0.786 sd 0.216 

Base case analysis

Base case analysis was conducted for all four randomized arms (e.g. four arms model). Secondary analyses comparing prednisolone versus NP and aciclovir versus no aciclovir were also conducted. For all analyses, cumulative mean costs were estimated for the 9 months follow-up period of the trial. All costs were expressed in 2006/2007 pounds sterling. The perspective of the analyses was that of the British National Health Service. Effectiveness was measured in terms of number of cases cured (e.g. House–Brackmann score = 1), and mean QALYs for the 9-month time horizon. As the time horizon for the analyses was less than a year, neither cost nor effectiveness outcomes were discounted. Incremental cost-effectiveness ratios (ICER) were calculated.

Sensitivity analysis

Deterministic and probabilistic sensitivity analyses were conducted. The latter involved attaching probability distributions to the model parameters and conducting Monte Carlo simulations (MCS). One thousand iterations were obtained for each MCS conducted. These MCS were used to produce cost-effectiveness acceptability curves (CEACs) (Fig. 2a) and cost-effectiveness scatter plots (Fig. 2b and 2c). In addition, the likelihood of an intervention being considered cost-effective (CE) for society's willingness to pay at threshold values of £10 000, £20 000, £30 000 and £50 000 were calculated (Table 5). CEACs for the two arms models are reported as Supplementary Data (available at Family Practice online).

F IGURE 2

(a) CEACs. Four arms model. These CEACs indicate that collectively the other interventions have only a 20% chance of being considered CE. (b) Incremental cost-effectiveness scatter plot. Prednisolone versus NP model. Scatter plot of the difference in cost and QALY pairs for the comparison of prednisolone compared with NP from the Monte Carlo simulation. A high proportion of the dots is allocated within the south east quadrant. Therefore, for those cases, prednisolone produced more QALYs and was less costly than NP and prednisolone is CE for these iterations. The opposite argument applies to those cases that fall within the north west quadrant (e.g. NP option is CE). Finally, for those iterations that fall within the north east and south west quadrants, the decision for or against prednisolone will depend on threshold value of willingness to pay for an extra QALY. (c) Incremental cost-effectiveness scatter plot. Aciclovir versus no aciclovir model. Scatter plot of the incremental cost and QALY pairs from the Monte Carlo simulation shows that the majority of the iterations lie within the north west quadrant (e.g. aciclovir more costly and less effective that no aciclovir)

F IGURE 2

(a) CEACs. Four arms model. These CEACs indicate that collectively the other interventions have only a 20% chance of being considered CE. (b) Incremental cost-effectiveness scatter plot. Prednisolone versus NP model. Scatter plot of the difference in cost and QALY pairs for the comparison of prednisolone compared with NP from the Monte Carlo simulation. A high proportion of the dots is allocated within the south east quadrant. Therefore, for those cases, prednisolone produced more QALYs and was less costly than NP and prednisolone is CE for these iterations. The opposite argument applies to those cases that fall within the north west quadrant (e.g. NP option is CE). Finally, for those iterations that fall within the north east and south west quadrants, the decision for or against prednisolone will depend on threshold value of willingness to pay for an extra QALY. (c) Incremental cost-effectiveness scatter plot. Aciclovir versus no aciclovir model. Scatter plot of the incremental cost and QALY pairs from the Monte Carlo simulation shows that the majority of the iterations lie within the north west quadrant (e.g. aciclovir more costly and less effective that no aciclovir)

Further sensitivity analyses related to changes in key parameters used in the model e.g. unit cost values or to changes in model assumptions relating to the derivation of cost and the definition of cure. Cost data are typically skewed to the right as there are usually a few trial participants for which costs are extremely high. A sensitivity analysis was conducted taking these potential outliers out of the analysis.

Potential drivers in these models are the probability of being cured or not cured at 3 months; therefore, threshold analysis was also used to explore the effect of the probability of being cured or not cured on the model results. In addition, subgroup analyses by age and sex were also performed. Finally, structural uncertainty was explored by assuming an exponential regression and gamma regression (together with gamma distributions for MCS) analyses for total costs instead of the original OLS regression.

Results

Comparison of all four randomized groups

On average, prednisolone only (PO) is the least costly and most effective of the four alternative interventions ( Table 5 ). Furthermore, it has approximately an 80% chance of being considered CE compared with the other treatments ( Fig. 2a and Table 5 ).

T ABLE 5

Cost-effectiveness results

Treatment Cost (£)  Cured cases a at 9 months (%)   ICER b QALYs  ICER c  Probability that intervention is CE for different threshold values for society's willingness to pay for a QALY (%)
 
£10 000 £20 000 £30 000 £50 000 
Four arms model          
    PO 182.34 85.6  0.719  79.1 77.4 76.9 75.9 
    Aciclovir + prednisolone 198.09 78.0 Dominated 0.718 Dominated 0.0 0.0 0.0 0.1 
    No treatment 205.14 78.0 Dominated 0.717 Dominated 12.5 9.5 7.2 5.2 
    Aciclovir only 219.62 78.0 Dominated 0.716 Dominated 8.4 13.1 15.9 18.8 
Prednisolone versus NP model          
    Prednisolone 231.98 94.4  0.718  79.3 77.5 77.0 76.0 
    NP 248.05 81.6 Dominated 0.717 Dominated 20.7 22.5 23.0 24.0 
Aciclovir versus no aciclovir model          
    No aciclovir 235.33 90.8  0.718  91.1 85.1 82.2 79.0 
    Aciclovir 246.63 85.4 Dominated 0.717 Dominated 8.9 14.9 17.8 21.0 
Treatment Cost (£)  Cured cases a at 9 months (%)   ICER b QALYs  ICER c  Probability that intervention is CE for different threshold values for society's willingness to pay for a QALY (%)
 
£10 000 £20 000 £30 000 £50 000 
Four arms model          
    PO 182.34 85.6  0.719  79.1 77.4 76.9 75.9 
    Aciclovir + prednisolone 198.09 78.0 Dominated 0.718 Dominated 0.0 0.0 0.0 0.1 
    No treatment 205.14 78.0 Dominated 0.717 Dominated 12.5 9.5 7.2 5.2 
    Aciclovir only 219.62 78.0 Dominated 0.716 Dominated 8.4 13.1 15.9 18.8 
Prednisolone versus NP model          
    Prednisolone 231.98 94.4  0.718  79.3 77.5 77.0 76.0 
    NP 248.05 81.6 Dominated 0.717 Dominated 20.7 22.5 23.0 24.0 
Aciclovir versus no aciclovir model          
    No aciclovir 235.33 90.8  0.718  91.1 85.1 82.2 79.0 
    Aciclovir 246.63 85.4 Dominated 0.717 Dominated 8.9 14.9 17.8 21.0 
a

Cured cases defined as House-Brackmann score = 1.

b

Incremental ICER using % cured cases.

c

Incremental ICER using QALYs.

Prednisolone versus NP model

When the proportion of cases cured (cost-effectiveness analysis) or QALYs (cost-utility analysis) is used as the measure of effectiveness, prednisolone has a lower mean cost and is more effective than the NP alternative ( Table 5 ). Thus, prednisolone dominates the NP intervention. Table 5 also shows that prednisolone is likely to be considered a CE treatment at all values for society's willingness to pay for a QALY. Figure 2b shows the cost-effectiveness scatter plot. The majority of the MCS cost-effectiveness result dots lay within the south east quadrant and for these prednisolone treatment is more effective and less costly than NP treatment.

Aciclovir versus no aciclovir model

Table 5 shows the incremental cost per case cured and per QALY for the comparison of aciclovir with no aciclovir. The no aciclovir alternative has on average lower costs and a higher proportion of individuals recovered. Therefore, on average, no aciclovir dominates aciclovir treatment. The probabilistic analysis reinforces this finding ( Table 5 and Fig. 2c ).

Sensitivity analysis

A wide range of sensitivity analyses were conducted. Results were not sensitive to the exclusion of the higher cost participants from the analysis, to halving or doubling the unit costs or when an exponential regression was used to estimate total cost for cured or not cured participants to allow the right skew for the cost data. PO appeared less likely to be considered CE when gamma regression and gamma distributions were used.

One-way sensitivity analyses were conducted on the difference in the probability of being cured at 3 months. The 95% confidence interval (CI) upper and lower limits for the difference in cure rates were used for this ( Table 1 ). Cost-effectiveness or cost-utility analyses results were not sensitive to these changes for prednisolone versus NP model.

However, results were sensitive to the probability of being cured at 3 months within the aciclovir versus no aciclovir model. When the difference in the probability of being cured at 3 months between the aciclovir arm and no aciclovir arm was 3.3% (the upper limit of the 95% CI), the ICER was £9576. Further threshold analyses were conducted and ICERs of about £20 000 and £30 000 were obtained for 2% and 1.5% differences in the absolute probability of cure, respectively. Therefore, the CI surrounding the difference in cure rates between aciclovir arm and no aciclovir arm is sufficiently wide to contain clinically and economically important differences.

Age group and gender

Regression analyses for total cost and for utility weights show age group variables as well as gender were statistically non-significant. Given these data, no estimates of incremental cost per QALY were estimated for different age groups or by gender.

Discussion

The results of the economic evaluation suggest that the use of prednisolone is likely to be considered CE. Aciclovir, in contrast, appears to be on average no more effective but more costly than no treatment or treatment with prednisolone. Thus, it is unlikely to be considered CE. The time horizon of the model was only 9 months. Therefore, an implicit assumption is that there are no further benefits and cost savings from the use of prednisolone after the end of the time horizon. Given the difference in cure rates that existed at 9 months, it is possible that should the time horizon be extended, treatment of BP with prednisolone would be associated with further gains in quality of life. Furthermore, it is likely that those who did not receive prednisolone would make more use of health services; thus, increasing their cost relative to those who received prednisolone.

The economic analysis used a modelling framework to estimate relative efficiency. This approach has the advantage of making the best use of the limited data available but it made the assumption that the main determinant of relative efficiency is whether or not the BP was cured or not. If a standard trial-based cost-effectiveness analysis had been conducted, it is likely that, on average, similar but less precised results would have been obtained. Furthermore, the lack of data on costs and the decision not to follow-up those deemed cured at 3 months would have necessitated similar assumptions being made in order to handle the missing cost and utilities data. 13

The data on costs used within the model came from a sample of only 74 of the trial participants. This led to a reduction in the precision of the estimates. Efforts were made to obtain data from more trial participants but these efforts were hampered by the fact that some general practices refused permission to view notes even though the participant had granted permission for their records to be reviewed. This appeared to be caused by uncertainty over whether the prior consent to view notes would still apply several months after initial recruitment of the participant and also the inconvenience of allowing investigators access to the practice. Despite this limitation, these data appear representative of the whole sample and the reasons for non-response were unconnected to the therapy the participant received or their outcomes (see 1a and b in the supplementary data , available in Family Practice online). With respect to the estimation of QALYs, measurements of health state utilities were censored for those trial participants who were judged to be cured at the 3-month follow-up. Therefore, an assumption was made within the modelling exercise that was tantamount to imputing utility data using the ‘last value carried forward’ method. Ordinarily, this approach while simple is normally considered to be a limited method of imputation. 14 , 15 However, in this situation, it may not be wholly unrealistic as these trial participants were judged to be cured at the time of censoring. Nevertheless, it assumes that there is no possible further improvement in health status for these people nor is there any possibility of relapse. This latter situation is clinically implausible unless there is an unrelated new episode of BP. The results of the economic evaluation would have been strengthened by further data on both costs and health state utilities.

Within the model, the results are driven by the probability of being cured at 3 months and to a lesser extent, the probability of being cured at 9 months. Both probabilistic and deterministic sensitivity analyses were conducted. The probabilistic sensitivity analysis focused on the statistical imprecision surrounding the model parameters using parameter distributions that were plausible and based upon the available data. Further deterministic sensitivity analysis was conducted to address uncertainty in the model structure or uncertainty surrounding model parameters that were obtained from outwit the RCT. The results of these sensitivity analyses indicate that conclusions are only sensitive to assumptions on the probability of being cured for the aciclovir versus no aciclovir model.

Conclusions

Overall, based on the data available, it appears that treatment of BP with prednisolone is likely to be considered CE while treatment with aciclovir is highly unlikely to be considered CE. Given the limited data available on costs and utilities, further data would be useful to confirm findings. Similarly, even though it is unlikely to change conclusion, further data on costs and outcomes in the longer term (i.e. for a follow-up greater than 9 months) would also serve to confirm the findings of the study.

Supplementary material

Supplementary Data can be found at http://fampra.oxfordjournals.org/

Declaration

Funding: National Institute for Health Research Health Technology Assessment programme (project number HTA 02/09/04).

Ethical approval: None.

Conflicts of interest: None.

We thank members of the BELLS Trial Group: Jillian M Morrison: Division of Community Based Sciences, University of Glasgow; Blair H Smith: Department of General Practice and Primary Care, University of Aberdeen; Brian McKinstry: Community Health Sciences, University of Edinburgh; Richard J Davenport: Department of Clinical Neurosciences, University of Edinburgh; Janet E Clarkson: Dental Health Services Research Unit, University of Dundee; Ken Stewart: St John's Hospital, NHS Lothian; Victoria Hammersley: Community Health Sciences, University of Edinburgh; Sima Hayavi: Division of Community Based Sciences, University of Glasgow; Anne McAteer: Department of General Practice and Primary Care, University of Aberdeen; Denise Gray: Department of General Practice and Primary Care, University of Aberdeen; Fergus Daly: Community Health Sciences, University of Dundee. We also thank Prof. John Cairns from London School of Hygiene and Tropical Medicine, University of London, who initially designed the economic component of the study and Dr Murray Smith from the Health Economics Research Unit, University of Aberdeen for useful comments. The Health Services Research Unit and the Health Economics Research Unit are both core funded by the Chief Scientist Office of the Scottish Governments Health Directorates. The views expressed in this report are those of the authors and not necessarily those of the funders. The paper was developed from a Health Technology Assessment on the early treatment with prednisolone or acyclovir in BP.

References

1
Petruzelli
GJ
Hirsch
BE
Bell's palsy
Postgrad Med
 , 
1991
, vol. 
90
 (pg. 
115
-
127
)
2
Bateman
DE
Facial palsy
Br J Hosp Med
 , 
1992
, vol. 
47
 (pg. 
430
-
431
)
3
Peitersen
E
Bell's palsy: the spontaneous course of 2,500 peripheral facial nerve palsies of different etiologies
Acta Otolaryngol Suppl
 , 
2002
, vol. 
549
 (pg. 
4
-
30
)
4
Neely
JG
Neufeld
PS
Defining functional limitation, disability, and societal limitations in patients with facial paresis: initial pilot questionnaire
Am J Otol
 , 
1996
, vol. 
17
 (pg. 
340
-
342
)
5
Salinas
RA
Alvarez
G
Ferreira
J
Corticosteroids for Bell's palsy (idiopathic facial paralysis)
Cochrane Database of Systematic Reviews 2004
  
Issue 4. Art. No.: CD001942. DOI: 10.1002/14651858.CD001942.pub2
6
Allen
D
Dunn
L
Aciclovir or valaciclovir for Bell's palsy (idiopathic facial paralysis)
Cochrane Database of Systematic Reviews 2004
  
Issue 3. Art. No.: CD001869. DOI: 10.1002/14651858.CD001869.pub2
7
Kirwan
JR
Power
LL
Glucocorticoids in Rheumatic Diseases
Reports on the Rheumatic Diseases Series 5
  
8
Sullivan
FM
Swan
IRC
Donnan
PT
, et al.  . 
Early treatment with prednisolone or acyclovir in Bell's palsy
N Engl J Med
 , 
2007
, vol. 
357
 (pg. 
1598
-
1607
)
9
British Medical Association & Royal Pharmaceutical Society of Great Britain.
 
British National Formulary. 2006. http://www.bnf.org/bnf (accessed May 2007)
10
Information Services Department for Scotland
NHS—National Services Scotland
  
http://www.isdscotland.org . (accessed May, 2007)
11
Curtis
L
Netten
A
Unit Cost of Health & Social Care
 , 
2006
UK
PSSRU. University of Kent
 
12
Furlong
WJ
Feeny
DH
Torrance
GW
Barr
RD
The health utilities index (HUI) system for assessing health-related quality of life in clinical studies
Ann Med
 , 
2001
, vol. 
33
 (pg. 
375
-
384
)
13
Manca
A
Palmer
S
Handling missing data in patient-level cost-effectiveness analysis alongside randomised clinical trials
Appl Health Econ Health Policy
 , 
2005
, vol. 
4
 (pg. 
65
-
75
)
14
White
IR
Wood
A
Royston
P
Editorial: multiple imputation in practice
Stat Methods Med Res
 , 
2007
, vol. 
16
 (pg. 
195
-
197
)
15
Kenward
MG
Carpenter
J
Multiple imputation: current perspectives
Stat Methods Med Res
 , 
2007
, vol. 
16
 (pg. 
199
-
218
)

Author notes

Hernández RA, Sullivan F, Donnan P, Swan I and Vale L, for the BELLS Trial Group. Economic evaluation of early administration of prednisolone and/or aciclovir for the treatment of Bell's palsy. Family Practice 2008; 26: 137–144.