Efficacy of Liposomal Bupivacaine and Bupivacaine Hydrochloride vs Bupivacaine Hydrochloride Alone as a Periarticular Anesthetic for Patients Undergoing Knee Replacement

This randomized clinical trial assesses the efficacy and cost-effectiveness of liposomal bupivacaine plus bupivacaine hydrochloride compared with bupivacaine hydrochloride alone for postoperative recovery and pain management among patients undergoing knee replacement.

Chiropractor: assumed to cost the same as physiotherapy session (£63) Osteopath: assumed to cost the same as physiotherapy session (£63) *including direct care staff costs and including qualification costs **including qualification costs

Intervention and index admission costs
The cost per vial of LB was £241.80 and each participant allocated to the intervention arm received a vial. This cost was provided by the manufacturer. Usual care (UC) arm (i.e. control arm) included use of 100 mg plain bupivacaine (diluted with normal saline) used for peri-articular infiltration. The full cost per vial was applied, even if participants did not receive the protocol stipulated dose because the medication is prepared for the surgery and cannot be re-used if only partially used.
No additional costs for normal saline or plain bupivacaine or use of needles / syringes were taken into consideration as these are minimal, were similar across both the study arms and were assumed to be included in the HRG code. The index hospital stay was converted into a Healthcare Resource Group (HRG) and valued using NHS Reference Costs (6). HRGs are groups of ICD-10 diagnoses and OPCS procedures which use comparable levels of healthcare resources. Based on previous research (6), we used the most common ICD-10 code reported in England for knee replacements: M179 (gonarthrosis, unspecified). We also used the most common OPCS procedure reported in that work that was associated with primary knee replacements: W401 (Primary total prosthetic replacement of knee joint using cement)(6) for total knee replacements, and W581 (Primary resurfacing arthroplasty of joint) and Z845 (Tibiofemoral joint, which the study Chief Investigator confirmed was the most common procedure for unicompartmental knee replacements [UKR] in SPAARK) for all UKRs. Both procedures resulted in the same HRGs (see Appendix 5: Table e2). In England, the large majority of index knee replacements consisted of a single finished consultant episode (FCE) and we assumed the same to apply in this trial. If participants reported post-op complications during their index hospital stay, we coded these as additional secondary ICD-10 diagnosis in the index FCE (finished consultant episode), resulting in differences in the CC scores between patients, which were reflected in the costing. This assumption was also based on the analysis of knee replacements in England (6).
Furthermore, we compared theatre time and length of stay during index admission as well as opioids taken after surgery in the two trial arms. We added cost differences to the HRG-based outlined above. This was done as follows: -Length of stay: we compared the length of stay of each participant with the trial average for the relevant procedure (total or unicompartmental knee replacement). Differences (in days) relative to the trial average were valued using the cost of an excess bed day; Note: the length of stay excluded days in intensive care or high dependency units, which were be costed separately.
-Theatre time: we compared theatre time of each participant with the trial average for the relevant procedure (total or unicompartmental knee replacement). Differences (in minutes) relative to the trial average were valued using the average cost per minute in theatre; -Opioids: we valued opioids taken during hospital stay and added these to the total hospital stay costs.

GP, community and outpatient visits
Participants were asked to record their attendances with a GP (surgery, home, or telephone), physiotherapist or occupational therapist (NHS or Private), outpatient clinic (NHS or Private), home care worker or social worker (NHS or Private) and other types of healthcare received as free text (these were allocated to existing resource use categories or additional categories where possible). Information on contacts with accidents and emergency departments were also collected. Participants were asked to report visits as a results of something to do with their knee. We costed all these visits using the national databases (see Appendix 5: Table e3).

Hospital admissions
Hospital admissions during the trial follow-up (i.e. admissions subsequent to the index admission) were costed based on the trial's readmission case report form. This was completed by research staff, and included ICD-10 codes for the hospitalisations, or, if these were not available, details on the reasons for admission, date of admission and discharge, time in intensive care and high dependency unit, and procedures performed during the intervention, were given. We used these data as the primary source for costing relevant admissions during the trial follow-up. Inpatient days beyond the trim point were costed as excess bed days. NHS Reference costs were applied (see Appendix 5: Table e3), and all admissions were assumed to have taken place in an NHS setting.
For hospital admissions recorded without sufficient information to attach an HRG, we valued these using weighted averages of admissions in the relevant trial arm by type (day case, emergency and planned admissions).

Medications
Patients were asked to report any medications taken, whether they were purchased or prescribed, and the dosage, duration, and frequency. Data on dosage, duration, and frequency is often missing. Each self-reported medication was categorised according to its chemical name where possible. Using all data from the 2019 Prescription Cost Analysis(4), the most common medication within each chemical name was identified, and the cost per item prescribed extracted (net ingredient cost per item).
Each medication was classified as likely to be one-off prescription or a long term prescription by physicians. For those drugs considered long-term, we identified the typically number of prescriptions per year based on recommended use and standard pack sizes from the British National Formulary(7).

Equipment and home changes
Participants were asked to report details of any changes to their home/equipment purchased or provided following the index surgery as well as their financial contribution to it. Self-reported equipment types were allocated into a category, and costs attached to each category (see Appendix 5:

Travel costs
Participants were asked to report costs of travelling to attend healthcare visits related to their knee.

Social services
Participants were asked to report details on use of social services (meals on wheels, home care, personal care assistant (£24 per hour), laundry services (£10.5 per service) and other), the number of weeks for which they used the service, and whether they were still using it. These are referred to as personal and social services (PSS).

Informal care
Participants were asked to report whether they received unpaid care from family or friends, the number of weeks for which any care was received, the number of hours of care provided per week, whether their carer took time away from paid work, and whether they were still receiving it. Minimum hourly wage (£8.21) were used if carer did not take time away from paid work. Mean hourly wage for all employee jobs (£18.03) were used to value the time of carers who took time away from paid work. Where data on weeks or hours per week of care were missing (but care is reported to have been received), mean imputation was used.

Lost productivity costs
Participants were asked to report whether they had to take time off paid employment due to their knee and, if so, how many days, if they lost pay and, if so, how much pay they have lost. Self-reported occupation were translated into a SOC category (https://onsdigital.github.io/dp-classification-tools/standard-occupationalclassification/ONS_SOC_occupation_coding_tool.html) and the days off work were costed using mean hourly wages for all employee jobs by SOC category (see Appendix 5: Table e5).
Participants were also asked to report how many days at work were affected by their knee and if so, to rank how much they were affected on a scale of 0 to 10. Participants not in paid employment were asked how many days did their knee affect their ability to carry out usual activities and, if so, to rank how much they were affected on a scale of 0 to 10. The self-reported rank was divided by 10 and multiplied by the number of days reported to ascertain the total days at work of lost productivity. These days were costed using mean hourly wages by SOC for those employed full-time and minimum hourly wage for those unemployed/retired. 10.03

Missing data
We followed best practice methods for addressing missing data in cost-effectiveness studies (8). Missing baseline data were imputed using unconditional mean imputation. Data on receipt of allocated interventions and deaths were considered to be complete, and so no imputation were performed. For components of resource use where participants provided responses to any questions in the resource diary, we imputed missing values as zero.
We used multiple imputation by chained equations to impute missing data on EQ-5D-5L utility scores, and cost components (except costs related to the allocated intervention), at each follow-up time point. Each missing value was imputed as a function of follow-up period, sex, age, recruitment site, and baseline EQ-5D score, updated EQ-5D score and components of costs, and the imputation model was run separately by randomised treatment. We used predictive mean matching to create a total number of 30 imputed datasets (i.e. the proportion of data missing across all time periods times 100). We imputed costs and EQ-5D-5L utility score in each period. In periods where death was observed we adjusted these. For costs, we assumed that they were incurred linearly over time, such that, if an individual died half way into a period, they incurred half the predicted costs.

Within-trial analysis
We reported descriptive statistics (means, SD as a minimum) for resource use, costs, and EQ-5D utilities at each follow-up time point using only complete data. Differences between arms for the EQ-5D-5L utilities were estimated using multi-level mixed effects linear regression models, to allow for multiple follow-ups clustered within participant. The model was adjusted for treatment allocation, an interaction between follow-up time and treatment allocation, recruitment site, and, in the case of EQ-5D, baseline utility score. Clustering by site was accounted for using robust standard errors, using the 'cluster' option in Stata. Other outcomes were analysed using simple regression models or t-tests, as appropriate.
Following multiple imputation, we estimated total costs and QALYs for all participants in the SPAARK study from the date of study recruitment until the earliest of death, from study, or the end of follow-up at 1 year. Our analysis followed intent-to-treat principles wherein healthcare resource use, costs and EQ-5D scores were analysed according to treatment allocation, regardless of the treatment actually received. We did not discount total costs and QALYs as the time horizon of the analysis is 12 months.
On each imputed dataset, we estimated mean costs (by type) and QALYs using separate analysis models, as described above. Estimates derived from each imputed dataset were combined using Rubin's rule to estimate the adjusted mean difference and standard error for each outcome. As a sensitivity analyses, we performed a complete case analysis, including only individuals who provided complete data over the 12 month trial duration.
We estimated the Incremental Cost Effectiveness Ratio (ICER) by dividing the mean cost difference between LB and usual care by the mean QALY difference.
We estimated the joint uncertainty around incremental total costs and QALYs (i.e. the difference between LB and usual care), and in the cost-effectiveness, by bootstrapping 1,000 times from each of the n imputed datasets (creating at least 30,000 bootstraps), running the estimation model on each bootstrapped dataset and extracting the estimated treatment effects. From these bootstrapped results, we calculated the probability that LB was more cost-effective than usual care for different threshold values per QALY gained (9). These were calculated by estimating the proportion of bootstrap replicates with a net monetary benefit (NMB) above 0 for each threshold value, where the NMB was given by the product of the mean difference in QALYs and the threshold value minus the mean difference in costs.      (min, max). Opioids were converted to oral morphine equivalent doses using standardized conversion tables 2 Summaries are mean (SD) 3 For participants discharged prior to day 3, it is assumed they took no opioids following discharge 4 For participants discharged prior to day 3, it is assumed they took the maximum daily dose of the opioids prescribed at discharge until discharge eTable 16. Analysis of OKS and AKSS Secondary Outcome    Differences between treatment arms are obtained from multilevel mixed-effects models, adjusted for baseline utility, type of surgery performed (total vs. partial knee replacement); robust standard errors were used to account for clustering by site; a time by treatment interaction was included in the model; the follow-up time point was used as a categorical variable.

2
Differences between treatment arms are obtained from regression models, adjusted for baseline utility, type of surgery performed (total vs. partial knee replacement); robust standard errors were used to account for clustering by site. 3 The QALY data for the baseline to six months period included EQ-5D-5L utility data collected at days zero, one, two, three and 42 days (6 weeks) post-operatively for participants who received their surgery. This is because these assessments varied in terms of their timing from randomisation. *data only expected for those who had surgery Missing baseline data were mean-imputed.  The interpretation is that, given a willingness to pay threshold of £20,000 per QALY gained, the probability that LB is cost-effective compared to control is 0.37 and from a NHS&PSS and a societal perspective, respectively. Abbreviations: LB -liposomal bupivacaine; PSSpersonal and social services; QALYquality adjusted life year