Volar locking plate versus external fixation for unstable dorsally displaced distal radius fractures–A 3-year cost-utility analysis

Aim To investigate the cost-effectiveness of Volar Locking Plate (VLP) compared to External Fixation (EF) for unstable dorsally displaced distal radius fractures in a 3-year perspective. Methods During 2009–2013, patients aged 50–74 years with an unstable dorsally displaced distal radius fracture were randomised to VLP or EF. Primary outcome was the incremental cost-effectiveness ratio (ICER) for VLP compared with EF. Data regarding health effects (Quality-adjusted life years, QALYs) was prospectively collected during the trial period until 3 years after surgery. Cost data was collected retrospectively for the same time period and included direct and indirect costs (production loss). Results One hundred and thirteen patients (VLP n = 58, EF n = 55) had complete data until 3 years and were used in the analysis. At one year, the VLP group had a mean incremental cost of 878 euros and a gain of 0.020 QALYs compared with the EF group, rendering an ICER of 43 900 euros per QALY. At three years, the VLP group had a mean incremental cost of 1 082 euros and a negative incremental effect of -0.005 QALYs compared to the EF group, which means that VLP was dominated by EF. The probability that VLP was cost-effective compared to EF at three years, was lower than 50% independent of the willingness to pay per QALY. Conclusion Three years after distal radius fracture surgery, VLP fixation resulted in higher costs and a smaller effect in QALYs compared to EF. Our results indicate that it is uncertain if VLP is a cost-effective treatment of unstable distal radius fractures compared to EF.


Introduction
The incidence of surgical treatment of distal radius fractures has increased since the introduction of the volar locking plate (VLP) at the turn of the 21st century [1]. VLP has become the most commonly used surgical method, while the use of percutaneous methods, i.e. percutaneous pinning or external fixation (EF), has been reported to decrease [1][2][3]. There is little evidence to support that any surgical method yields superior clinical outcome as compared to others for treatment of distal radius fractures [4][5][6]. Other factors than final clinical outcome may therefore be allowed to influence treatment method choices. In a setting with limited health care resources, cost-effectiveness of different methods may be an important aspect to address in the choice of treatment, i.e. if the surgical methods have a reasonable incremental cost in relation to their effects. There is some evidence suggesting that VLP is not a cost-effective surgical technique when compared to percutaneous pinning [7,8]. To the best of our knowledge, health economic assessments of other treatment methods for distal radius fractures are largely lacking. No study has investigated cost-effectiveness of distal radius fracture surgery beyond a one-year perspective. The purpose of this study was to assess the cost-effectiveness of VLP versus EF for surgical treatment of patients 50-74 years old with a dorsally displaced distal radius fractures during the first 3 years after distal radius fracture surgery.

Materials and methods
This study is a cost-utility analysis based on patients included in a previously published randomised controlled trial (RCT) comparing VLP with EF regarding functional outcome [6,9]. Patients eligible were 50-74 years of age with a distal radius fracture of >20 degrees dorsal angulation after a low energetic trauma presenting at a second-level trauma hospital in Stockholm, Sweden, during September 2009 to February 2013. Full inclusion and exclusion criteria are presented in Table 1. 140 patients were randomised through opening of sealed opaque envelopes to EF (Hoffman Compact T2, Stryker, Switzerland) or VLP fixation (2.4 mm Variable Angle LCP Two-Column Volar Distal Radius Plate, Synthes, Switzerland). Data regarding health effect was prospectively collected during the trial period. Cost data was collected retrospectively. The clinical 1-year and 3-year results [6,9] displayed no differences in Patient-Reported Outcome Measures (PROM) after the first 3 months. The analysis was conducted on an intention-to-treat basis. A power calculation was performed for detection of a 10-points difference in the main outcome Disability of the Arm, Shoulder and Hand (DASH) at one year follow-up of the initial RCT. A separate power calculation for the cost-utility outcomes was not conducted.

Cost-effectiveness
This cost-utility analysis has been conducted using a health care perspective as well as using a broader perspective including production loss. The time horizons used are one and three years. The primary outcome was the incremental cost-effectiveness ratio (ICER) for VLP level are available in the publication. However, individual clinical data are only available in the statistical files. The ethical permit does not allow us to make other than group data freely available online, but all files are fully available on request from the corresponding author, or from the Institution for Clinical Research and Education, Sodersjukhuset hospital, Karolinska Institutet, Sweden. We confirm that other researchers will be able to access the data in the same manner as the authors, since we are in posession of the statistical files without restrictions others than stated above. All other data is presented in its complete form in the submitted paper and its tables/figures/ appendices. Data requests may be sent to the corresponding author (cecilia.mellstrand-navarro@sll.se) or the Karolinska Institutet telephone:0+46 8 524 875 04, E-mail:per. tornvall@ki.se compared with EF using a health care perspective including production loss. The ICER was defined as the difference in mean total cost per patient divided by the difference in mean Quality-adjusted-life years (QALY) per patient, expressed as the incremental cost per gained QALY for VLP compared with EF. If the mean difference in QALYs was negative and the mean difference in total cost positive, no ICER was calculated, as VLP then was considered to be dominated by EF. If the mean difference in total cost was negative and the mean difference in QALYs positive, no ICER was calculated, as VLP then was considered to dominate EF. Both costs and QALYs were in line with national guidelines discounted at a discount rate of three percent [10].

Costs
Total costs per patient were calculated by first identifying and estimating the resource use associated with each surgical method and then valuing each resource using the unit costs presented in Table 2. The direct costs and indirect costs during the first year and up to 3 years for each treatment were summed up.

Unit costs
All unit costs are presented in Table 2. Unit costs for operating theatre including staff were derived from a report by the Swedish Agency for Health Technology Assessment and Assessment of Social Services (SBU) [11]. Costs regarding in-and outpatient care including emergency ward visits were collected from the diagnose-related group (DRG [12]) financial reimbursement system used at the hospital. Costs for drug usage were calculated from prices defined in FASS [13] (a compilation from the pharmaceutical industry with information about drugs used in Sweden) for a Defined Daily Dose as defined in the Drug Registry of the Swedish National Board of Health and Welfare [14]. Unit costs for reoperations were calculated based on estimations by the study group regarding surgical time and material usage. All costs above are considered as direct costs. Indirect costs consisted of production loss due to sick leave after the surgery. The unit cost regarding production loss per day was derived from Statistics Sweden [15], using the mean income for adults 20-74 years plus taxes and social services fee. Costs from 2016 were converted to 2018 years prices using a 2% mark-up for every year. All costs are presented in euros converted from Swedish kronor (SEK) with an exchange rate of 0.0978.

PLOS ONE
Dorsally displaced distal radius fractures -A 3-year cost-utility analysis

PLOS ONE
Dorsally displaced distal radius fractures -A 3-year cost-utility analysis

Resource use
All resources needed for each treatment method were identified by the research group. Resource use data for surgical time for the primary surgery was derived from prospectively inserted data in the surgery software system used at the hospital (Orbit [16]). Inpatient and outpatient visits for diagnoses related to the initial injury and any possible related complication (International Classifications of Disease, ICD-10 codes [17] specified in Appendix) were retrieved at an individual level as registry data from the Swedish National Board of Health and Welfare. Drug usage was defined as prescription of antibiotics and analgesics (Anatomical Therapeutic Chemical Classification [18], ATC drug codes specified in Appendix) collected as registry data from the Swedish National Board of Health and Welfare. Data regarding sick leave for diagnoses related to the initial injury and any possible related complication (ICD-10 codes specified in Appendix) were collected as registry data from the Swedish Social Insurance Agency. Any reoperations were detected by search of patient records, and/or registry data retrieval from the Swedish National Board of Health and Welfare regarding surgical procedures related to any possible related complication (NOMESCO classification for surgical procedures codes [19] specified in Appendix). Estimations of resource use were performed by the study group for occupational therapy and x-rays since no complete registry or study protocol source was available. The time frame for all resource use was set to from the date of the injury to the date of the 3-year follow-up.

Effectiveness
Effectiveness of treatment was estimated using Quality of Life Adjusted Life Years (QALYs) [20,21]. QALYs are a composite measure of survival and Health related Quality of Life, HRQoL. One QALY can be interpreted as the equivalent of one year in full health. The QALYs following each treatment during the study period was calculated on an individual level using the area under the curve (AUC) approach [22]. QALYs for each time interval were calculated by taking the average of the HRQoL at two adjacent time points multiplied with the time in years spent in each time interval. The QALYs gained at 1 year and at 3 years were summarized and an average for each time period was calculated. The HRQoL of the patients was estimated using EuroQol 5 dimensions, EQ-5D-3L [23] and was reported by trial participants at baseline, 2 weeks, 6 weeks, 3 months, 1 year and 3 years postoperatively. EQ-5D-3L is a measure of health status and consists of a questionnaire with five questions and a visual analogue scale (EQ-VAS) [23]. The five questions each represent a dimension of health; mobility, self-care, usual activities, pain/discomfort and anxiety/depression. Each question has three response levels and can be combined into a health profile of five digits [23], which was converted into a health state value using a value set from the United Kingdom (UK) [24].

Statistical analyses
Data were analysed using SPSS version 26. A complete case analysis was conducted to avoid violating the assumption that data was missing at random, i.e. no imputations were made and only participants with complete data were analysed.
Categorical data was compared with Chi-square test. Normality was tested with Shapiro-Wilks test for all continuous variables. For normally distributed variables independent Student's t-test was used. Skewed distributed data was compared with Mann-Whitney U-test. Kolmogorov-Smirnov's test and Kruskal-Wallis´test were used to confirm statistical significance for non-parametric comparisons. The level of statistical significance was set to p<0.05 in twosided tests. Linear regression was used to adjust mean differential QALYs at 1 year and 3 years for imbalance between groups in EQ-5D-3L index scores at baseline [25]. Health state values (the EQ-5D-3L index scores in this study) at baseline (before treatment) is often invariably imbalanced between trial arms and it is recommended that the comparisons between treatments are adjusted for this imbalance as it otherwise will contribute to a difference in QALYs that is not an effect of the treatments [25]. Therefore, the difference in mean QALYs between VLP and EF was adjusted for differences in EQ-5D-3L index scores between VLP and EF at baseline (before surgery).
The non-parametric bootstrapping approach with replacement [26] was used to determine the level of sampling uncertainty around the ICER. The bootstrap was performed as a resampling from the original sample to create 1000 random samples. In each bootstrap sample, 58 individuals among the VLP patients and 55 individuals among the EF patients were randomly selected with equal probability and with replacement after each individual selection. To adjust for baseline differences in EQ-5D-3L index scores between the groups [25], we calculated the adjusted differential QALYs (VLP = intervention, EF = control) in each sample.
1000 estimates of incremental costs and effects were generated. The bootstrap is presented in a cost-effectiveness plane [26]. From the bootstrap a cost-effectiveness acceptability curve (CEAC) was derived at 1 year and 3 years, to express the probability that VLP is cost-effective in comparison to EF for a range of thresholds for willingness to pay (WTP) per gained QALY [26].
A threshold of 35000 euros was chosen as maximum WTP per gained QALY, which approximates the 30000 UK pounds sterling used by the National Institute for Health and Clinical Excellence (NICE) in the UK [27].

Ethics
The conduction of this study was approved by the Regional board for ethical vetting, Stock-

Results
Of the 140 patients randomised, 6 dropped out before the first year and 16 thereafter, leaving 118 patients for the 3-year follow-up. There were no missing data regarding resource use. Of the 118 patients, five had not filled in all EQ-5D-3L questionnaires and were excluded, leaving 113 patients (VLP n = 58, EF n = 55) for the cost-effectiveness analysis. From the EF group, four patients were converted to VLP intraoperatively and four received a volar plate within the first two weeks after primary surgery, but they were still evaluated within the EF group. Baseline characteristics are presented in Table 3.

Resource utilization and costs
Resource utilization is presented in Table 4. All costs are presented in Table 5 and Fig 1. The mean total cost was significantly higher for the VLP group compared with the EF group at 1 year (mean difference; MD: 878 euros, p = 0.006). Mean total cost increased for both groups until the 3-year follow-up, and VLP costs were still significantly higher (MD: 1 082 euros, p = 0.012).

Health-related quality of life
EQ-5D-3L index scores are presented in Table 6 and Fig 2. At 2 and 6 weeks, the VLP group had statistically significant better EQ-5D-3L index scores than the EF group, but differences did not remain at later follow-up time points. Mean EQ-5D-3L index scores improved continuously between all timepoints but was still lower than pre-injury levels at 3 years. Mean total QALYs during the first year was 0.814 in the VLP group and 0.787 in the EF group (p = 0.236) ( Table 7). After adjustments for baseline differences in EQ-5D-3L index scores between the groups, the difference in mean total QALYs was 0.020 (p = 0.344) in favor of the VLP group. At 3 years, mean total QALYs was 2.530 in the VLP group and 2.518 in the EF group (p = 0.852). The adjusted mean difference was 0.005 (p = 0.932) in favor of the EF group (Table 7).

Cost-utility analysis
From a health care perspective, the ICER at 1 year was 22 100 euros per QALY for VLP fixation compared to EF (Table 7). When including production loss, the ICER increased to 43 900 euros per QALY. At 3 years, VLP resulted in higher costs and a smaller effect in QALYs than EF, independent of whether production loss was included or not. This means that VLP was dominated by EF in the longer time horizon. The bootstrap analyses of the estimates including production loss are presented in cost-effectiveness planes (Fig 3). The scatterplot covers all four quadrants indicating uncertainty about whether or not VLP is cost-effective and at what value it is cost-effective compared to EF. The Cost Effectiveness Acceptability curves (CEAC) in Figs 4 and 5 summarize the probability of VLP being cost-effective compared to EF at one and three years respectively. At a willingness to pay threshold of 35 000 euros per QALY, the probability that VLP is cost-effective compared to EF is around 50% at 1 year and 40% at 3 years. At 3 years, the probability that VLP is cost-effective does not exceed 50% independent of the willingness to pay per QALY.

Discussion
The purpose of this study was to assess the cost-effectiveness of VLP compared to EF. The study shows that at 3 years, VLP patients had higher costs and a smaller effect (although not statistically significant) in QALYs compared to EF patients independent of the perspective

PLOS ONE
Dorsally displaced distal radius fractures -A 3-year cost-utility analysis used, indicating that VLP is not cost-effective compared to EF. At 1 year, the VLP group had gained more QALYs than the EF group, and the incremental total cost per QALY gained for the health care perspective (excluding production loss) was below the threshold recommended by NICE. However, when including production loss, the threshold was exceeded. Between 1 year and 3 years, VLP patients increased their costs more than EF patients and EF patients increased their EQ-5D-3L index scores more than VLP patients. The statistical analyses displayed a high level of uncertainty surrounding the ICER, which implies that further studies are needed to support our findings. There are no previous cost-utility studies comparing VLP with EF, but there are some studies comparing VLP with percutaneous pinning. Tubeuf et al [8] found a statistically significant incremental cost of 815 euros (converted from UK pounds sterling) after one year for VLP compared with percutaneous pinning. As VLP patients had a smaller gain in QALYs (0.008) than in our study, the resulting ICER was higher (100 295 euros per QALY). However, they did not investigate patients beyond the first year. Karantana et al [7] presented a study comparing VLP with percutaneous pinning and optional EF (11/64 patients) and showed a statistically significant incremental cost of 801 euros (converted from UK pounds sterling) after 1 year. They also presented a smaller gain in QALYs (0.0178) for VLP patients than our study, resulting in an ICER of 44 990 euros per QALY for the VLP group in comparison with the percutaneous pinning group.

PLOS ONE
Dorsally displaced distal radius fractures -A 3-year cost-utility analysis

PLOS ONE
Dorsally displaced distal radius fractures -A 3-year cost-utility analysis

PLOS ONE
Dorsally displaced distal radius fractures -A 3-year cost-utility analysis Differences in EQ-5D-3L index scores and resulting QALYs were very small in the studies of Tubeuf [8] and Karantana [7], which is in accordance with the findings in our study. Even small differences in total costs render large differences in ICER due to small differences in

PLOS ONE
QALYs. As VLP was associated with higher costs, VLP would still not be considered cost-effective even if there were no differences in QALYs.
The major strength of the present study is the relatively long follow-up period as treatment-related costs still occur after the first year, and HRQoL continues to improve. Another strength is that the study is conducted within the scope of a randomised trial, thus decreasing the risk of an impact on the results of potential biases. One strength is also that we have used registry data from the Swedish National Board of Health and Welfare, thereby capturing any resource use occurring at other hospitals or care providers. The use of registry data is also a limitation as we searched for ICD-10 codes and drug prescriptions that we assumed could be associated with the distal radius fracture, possibly rendering an overestimation of outpatient visits and drug usage. Another limitation is that we could not, in the retrospective perspective, evaluate the resource use of occupational therapy and x-rays and therefore had to make an estimation. Moreover, there was no data on primary care or nurse visits. Lastly, the study population is relatively small, thus limiting the power of detecting small differences between groups.
In conclusion, VLP fixation was associated with higher costs and resulted in fewer QALYs gained compared to EF at 3 years after distal radius fracture surgery., At this time horizon, the probability of VLP being cost-effective as compared to EF did not exceed 50% when including production loss, independent of the willingness to pay per QALY when adaption a perspective including production loss. Thus, our results indicate that it is uncertain if VLP is a cost-effective treatment of unstable distal radius fractures compared to EF.