The detailed study protocol was published earlier (4). The period of study was six months from the date of surgery for the economic evaluation.
Design, setting and participants
A pragmatic multi-centre three arm (parallel groups) RCT was done in accordance to Consolidated Standards of Reporting Trials (CONSORT) (additional file 1) [9]. The economic evaluation was done according to the CHEERS checklist (additional file 2) [10]. The study was done in the professorial gynaecology unit of the North Colombo Teaching Hospital, Ragama and the gynaecology unit of the District General Hospital, Mannar, Sri Lanka. Eligible participants were patients requiring hysterectomy for non-malignant uterine causes. Exclusion criteria was uterus>14 weeks, previous pelvic surgery, any medical illness which caution/contraindicate laparoscopic surgery, and those requiring incontinence surgery or pelvic floor surgery. Eligible patients were aware that they would be randomly assigned to undergo one of the three procedures. The main exposure variables were NDVH and TLH. The control group consisted of patients undergoing TAH (see Fig. 1). Standard techniques of TLH, TAH and NDVH was followed and the decision to convert to laparotomy at TLH or NDVH was at the discretion of the primary surgeon [11,12].
Sample size calculation
A difference of seven days was considered as clinically significant based on a study published in the BJOG by Ottosen et al [13]. A sample size of 49 patients per arm was required considering a type I error of 0.01 and a power of 80% and with a possible loss to follow up rate of 20% to detect a seven day difference between the three types; TLH versus TAH versus NDVH. The reason for a type 1 error of 0.01 (two-sided) was for a Bonferroni adjustment for multiplicity to keep the overall p value to less than 0.05 after pairwise comparisons.
Randomization
Block randomization in multiples of three was done at each study site by opening sealed envelopes containing computer-generated block randomization numbers, with block sizes of six and nine to ensure roughly equal numbers of patients in each arm at any point in the study. The patients and medical team were not blinded to the intervention.
Outcome measures
The primary outcome measure was time to recover following hysterectomy (earliest time to resume all or a combination of activities done prior to surgery; resumption of cooking, washing clothes, sexual activity and occupation). The secondary outcomes were operative time, time under anaesthesia, post-operative hospital stay, blood loss, pain scores, complications, quality adjusted life years (QALYs) and cost per procedure (direct hospital related costs and indirect patient borne costs).
Complications were graded and converted to an ordinal scale as follows; no complication (1), postoperative fever (2), skin wound infection (2), urinary tract infection (2), blood transfusion (2), laparotomy (3), bowel damage (4), bladder injury (4), ureteric injury (4), fistulae (4). If a patient had more than one complication the total score of all the complications were taken.
Quality Adjusted Life Years (QALYs) were estimated based on the EuroQol—Five Dimension Three Level (EQ-5D-3L) Questionnaire sequentially at pre-operative baseline, post-operative day two, seven, 6-weeks, 3-months and 6-months [14]. The EQ-5D-3L is a generic patient reported outcome measurement tool on health-related quality of life with five dimensions (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression) and three self-rated response levels of severity (e.g. no problem, some problem, major problem) for each dimension. Sri Lankan EQ-5D-3L weights were applied to estimate utility values at each of these time points [15]. Within trial QALYs were calculated for each route (TAH, NDVH and TLH) using area under the curve (AUC) estimated by integral calculation for each of the follow-up periods respectively.
Data analysis
All patients were included in the analyses. Data analysis was based on the intention-to-treat principle. A Kaplan Meier survival analysis was done for the primary outcome which is time to recover following hysterectomy. The end-point was the actual time to recover. Post hoc pairwise comparison between TAH, NDVH and TLH was done using COX regression. The other outcomes; operative time, anaesthesia time, post-operative hospital stay, cost, pain scores and blood loss were analysed using Kruskal-Wallis test with post hoc comparisons. A fully specified statistical analysis plan and database is provided (additional files 3 and 4).
Economic evaluation
The economic evaluation took a societal perspective and included direct costs related to utilization of healthcare resources in the hospital and indirect costs borne by the patient. A micro-costing approach was adopted to calculate utilization of hospital resources from the time of presentation to the gynaecology clinic up to six months after surgery. A detailed explanation of the methodology used in calculating costs was also published [16].
The incremental cost-effectiveness ratio (ICER) was estimated using median cost for the intervention arms (TLH or NDVH) minus the median cost for the standard treatment (TAH) arm divided by the median time to recover of the intervention arms (TLH or NDVH) minus the median time to recover of the standard treatment (TAH) arm. The ICER for the worst case scenario was calculated using third quartile (Q3) of time to recover and costs of the intervention arms (TLH and NDVH) versus the standard treatment (TAH). The ICER for the best case scenario was calculated using first quartile (Q1) of time to recover and costs of the intervention arms (TLH and NDVH) versus the standard treatment (TAH).
The incremental cost-utility ratio (ICUR) was estimated using mean direct cost for the intervention arm (TLH or NDVH) minus the mean direct cost for the standard treatment arm (TAH) divided by the QALYs of the intervention arm (TLH or NDVH) minus the QALYs of the standard treatment arm (TAH). The willingness to pay (WTP) threshold was considered to be 25% of the GDP per capita income [17]. Therefore a lower estimate of USD 1000 was considered based on the GDP per capita income of USD 4065 for the year 2017 [18]. The net monetary benefit (NMB) was considered as the incremental QALYs multiplied by the WTP threshold minus the mean incremental costs of the intervention group compared to the standard management group.
All cases were included in the analysis as there was a likelihood of either overestimating or underestimating costs if an uncomplicated or a complicated case was excluded. Missing data was not excluded and multiple imputation was used for missing data which was less than 5%. The costs are reported in USD (Exchange rate, 1 USD= LKR. 144.9625, as at 01/01/2017).
A deterministic sensitivity analysis was done to assess the robustness of costing assumptions. The worst case scenario considered a 5-year shelf life for surgical instruments, an additional 30 minutes of operating time, an additional post-operative day, doubling of the cost of complications and readmissions, doubling of utility costs and discounting rate of 10% using a standard discount table [19]. The best case scenario considered a 5-year shelf life for surgical instruments, 30 minutes less of operating time, a hospital stay of one day less post-operatively, with no complications and readmissions, and discounting rate of 10%.