Methods

Parts of this chapter are reproduced from Worthington et al.1 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.

Parts of this chapter are reproduced from Hashim et al.2 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.

Trial design

The UNBLOCS trial is a multicentre, pragmatic, randomised controlled, parallel-group trial of ThuVARP versus standard TURP. The trial was conducted in four university teaching hospitals and three district general hospitals in the UK, and recruited patients from June 2014 to December 2016. Men with BPO who were suitable for surgery were randomised to one of the two trial surgical treatment arms. The main aim of the study was to establish equivalence in clinical improvement, 12 months after the two surgical techniques. Both clinical and patient-reported outcomes were used to demonstrate equivalence: Qmax and IPSS.

Ethics approval and research governance

Approval from the National Research Ethics Service Committee South Central – Hampshire B Ethics Committee (13/SC/0644) was received on 15 January 2014 and applied to all NHS sites who took part in the study.

The trial is registered at the International Standard Randomised Controlled Trial Number registry with the reference ISRCTN00788389.

All patients provided their written informed consent to participate before entering the study. Consent was taken by the consultant, a dedicated research nurse or a designated team member. All adverse events were recorded, and serious adverse events that were categorised as unexpected and related were notified to the sponsor and Research Ethics Committee within 15 days.

A number of minor protocol changes were made during the trial.1 The secondary outcome measures were reviewed and updated to include:

• post-void residual urine at 12 months post surgery
• change in haemoglobin and serum sodium levels postoperatively as measures of blood loss and the absorption of irrigation fluid
• postoperative catheterisation time.

These measures were added to broaden the picture of surgical outcomes reported and were prespecified in the statistical analysis plan.18 The protocol and study documentation were also updated during the trial to make clarifications to study procedures, particularly for catheterised participants, and improve the formatting of health economic questionnaires, as well as making other adjustments to procedures to improve questionnaire return, and the addition of one other study site.

Participants

Patients presenting with either bothersome LUTS or urinary retention (secondary to BPO), who were suitable for TURP surgery, were recruited to the trial. Patients were excluded if they met any of the following criteria:

• had neurogenic LUTS
• had a prostate cancer diagnosis, or a prostate-specific antigen test outside the normal range without prostate cancer excluded
• had undergone previous prostate or urethral surgery
• were unable to give informed consent or complete trial documentation.

Interventions

Patients were randomised in a 1 : 1 ratio to receive either TURP or ThuVARP. ThuVARP uses laser technology to vaporise and resect the prostate, whereas TURP uses an electric current to resect the prostate. The procedures are otherwise similar in that a cystoscope (telescope) goes through the urethra into the bladder and irrigation fluid is used (saline or glycine). The prostate is resected in small ‘chips’ and these fragments are then evacuated with a device called an ‘Ellick’. The procedures are therefore minimally invasive, with no incisions made in the skin, and utilise the urethral meatus to enter the body and reach the prostate. During ThuVARP, everyone in the operating room needs to wear special goggles to protect the eyes. In this trial, all surgeons conducting the procedures were at consultant level, and mentored by the chief investigator or another principal investigator already certified as competent in the ThuVARP technique, as described in Surgeon training in using the laser. Given the pragmatic nature of the study, centres continued to follow their usual practices (e.g. monopolar or bipolar TURP). Concomitant procedures during BPO surgery were allowed and details of these were recorded.

It should be noted that, owing to the trial design, surgeons needed to list patients as either day case or inpatient before randomisation, which took place when the patient was already anaesthetised. Two centres listed the majority of both their TURP and laser cases as day-case procedures, whereas the remaining five sites listed all TURP and laser cases as inpatient procedures.

Surgeon training in using the laser

ThuVARP essentially uses the same surgical skill set as that for the TURP procedure, which is part of core clinical practice for all urologists in the UK, including the trial surgeons who performed both procedures. The experience of the chief investigator and of other urologists both in the UK and Europe indicates that a maximum of 15 ThuVARP laser cases can assure a surgeon’s competence in the ThuVARP laser procedure.

All surgeons were mentored by the chief investigator or another principal investigator already certified as competent in the ThuVARP technique, and certified by an independent assessor, using standard criteria, before the official study commenced. First, surgeons observed the chief investigator or principal investigator performing one or two procedures. The chief investigator or principal investigator then observed the surgeons in each centre perform 2–5 procedures during site visits. The surgeons then performed 5–10 procedures without supervision, following their own trust’s clinical governance and audit guidelines. Competency was assessed by an independent assessor using the Intercollegiate Surgical Curriculum Programme work-based assessment. If competency had not been achieved by this stage, then further procedures would have been observed and further training would have been provided by the chief investigator until the competency criteria were met. However, in practice, all surgeons were signed off as competent at their first assessment.1

Outcome measures

Primary outcome

International Prostate Symptom Score

The IPSS, a well-established and validated patient-reported outcome, was collected at 12 months. Participants filled in a questionnaire concerning their LUTS, which produced a score from 0 to 35, with higher scores indicating more severe symptoms.

Maximum urine flow rate (ml/second)

The Qmax was measured at participants’ 12-month follow-up assessment.

These measures were not collected if the participant had an indwelling catheter or was conducting intermittent catheterisation, if he was unable to void.

These outcomes address the primary research question for the trial. The IPSS and Qmax are internationally accepted, and are the most frequently used primary outcomes in BPO studies, thereby making results from this study comparable with those from others. No core outcome measures for BPO are listed on the COMET (Core Outcome Measures in Effectiveness Trials) Initiative website.

Secondary outcomes

Surgical complications

Any surgical complications were recorded and graded using the Clavien–Dindo classification.19 The number of complications experienced per participant was explored, along with the worst event per participant. There are five gradings:

1. deviation from normal postoperative course (e.g. analgesics) but not requiring pharmaceutical or surgical intervention
2. requiring pharmaceutical treatment
3. requiring surgical, endoscopic or radiological intervention (3a, not under general anaesthetic; 3b, under general anaesthetic)
4. life-threatening complication that requires intensive care
5. death of the participant.

Surgical outcomes

The length of hospital stay was measured as the number of days that the participant remained an inpatient in hospital after the procedure. The binary blood transfusion rate, time to successful trial without catheter, catheter use, post-void residual urine, blood loss during surgery (change in haemoglobin level) and absorption of irrigation fluid (change in serum sodium) were also collected for each arm.

Trial participants were asked to complete the following instruments.

International Consultation on Incontinence Questionnaire – Male Lower Urinary Tract Symptoms (ICIQ-MLUTS)

Both a voiding and incontinence score were generated from this questionnaire, as were daytime and night-time voiding frequency data.

International Consultation on Incontinence Questionnaire – Male Sexual Matters associated with Lower Urinary Tract Symptoms (ICIQ-MLUTSsex), and International Index of Erectile Function-5 (IIEF-5)

An overall erectile dysfunction score was generated from the IIEF-5 questionnaire, and the ICIQ-MLUTSsex questionnaire provided data on erection and ejaculation quality and how bothersome these aspects were for the participant.

International Consultation on Incontinence Questionnaire – Lower Urinary Tract Symptoms Quality of Life (ICIQ-LUTSqol)

This is a 20-item questionnaire about patients’ urinary problems, based on the King’s Health Questionnaire. Subscores were created for role limitations, physical implications, social limitations, personal relationships, emotions, sleep/energy and severity measures.

International Consultation on Incontinence Questionnaire – Satisfaction (ICIQ-satisfaction)

This is a 20-item questionnaire on patients’ satisfaction with surgery. Overall satisfaction (0–10) and pain suffered were explored.

Equivalence margin

In the light of the existing literature, TURP and ThuVARP were expected to be broadly similar in terms of the primary outcomes, with potentially more differences in the secondary outcomes.13 Therefore, an equivalence trial was chosen to allow greater focus on CIs, so that any clinically important difference would not be ruled out.

The sample size calculation was based on the consideration that patients randomised to the ThuVARP arm should have clinical outcomes that are equivalent to the outcomes of those patients who are randomised to the TURP arm. For the primary outcomes, a difference in LUTS score of no more than 2.5 points on the IPSS and of 4 ml/second in Qmax were hypothesised as suggesting equivalence. These were deemed appropriate for the following reasons:

• The minimal clinically important difference for the IPSS is generally accepted to be 3 points;13,20 however, a previous trial of ThuVARP versus TURP used a minimal clinically important difference of 2 points. The team felt that a level between these would be more suitable.
• There is no minimal clinically important difference in flow rate that is accepted in the literature; however, 2 ml/second has been quoted previously.8
• Discussions between clinicians, both in the trial team and with other urologists, led to an overall consensus that differences of no more than 4 ml/second in Qmax and of 2.5 points on the IPSS would be deemed clinically equivalent.

Sample size

This study is powered to establish equivalence in clinical improvement (measured with the IPSS and Qmax). A Chinese trial13 observed differences of 0.4 ml/second (95% CI –2.0 to 2.8 ml/second) in Qmax and 0.4 points (95% CI –0.7 to 1.5 points) in the IPSS between ThuVARP and TURP. Variability (standard deviation; SD) in data at 12 months was approximately 6.0 ml/second (Qmax) and 3.0 units (IPSS), but previous trials of TURP have reported greater variability of around 9 ml/second (Qmax) and 5 units (IPSS).21,22

After considerable discussions between clinicians both inside and outside the trial, we specified differences of 4 ml/second in Qmax and 2.5 units in IPSS as demonstrating equivalence. Equivalence studies often use an alternative hypothesis of a difference of zero between treatments. However, the Chinese trial observed differences of around 0.4 ml/second and 0.4 units for Qmax and IPSS, respectively. Incorporating these as alternative hypotheses ensured adequate power to demonstrate equivalence if the treatments were indeed similar but not identical.

Assuming SDs of 9 ml/second for Qmax and 5 units for IPSS, the target sample size needed to complete the 12-month follow-up was 163 participants in each arm. Using nQuery Advisor (Statistical Solutions Ltd, Cork, Ireland), this provided 85% power to demonstrate equivalence for Qmax and just over 90% power for IPSS, at a two-sided alpha of 5%. Assuming 20% loss to follow-up following randomisation, it would be necessary to recruit 410 participants in total. This loss to follow-up was a conservative estimate from our experience of previous trials.1

Randomisation and implementation

Men identified as eligible by a consultant or a research nurse were introduced to the trial at their clinical appointment or over the telephone. Eligible men who consented to take part in the trial were referred for LUTS surgery, if they had not already been listed. Randomisation took place in the anaesthetic room once the patient had been anaesthetised. An automated web/telephone randomisation system, provided by the Bristol Randomised Trials Collaboration, was used to randomly allocate patients to ThuVARP or TURP. Randomisation was stratified by centre and patient eligibility classification at baseline (bothersome LUTS or urinary retention). Patients were randomised in random blocks of two, four and six.

Blinding

One of the main strengths of this study was the successful blinding of participants, which is unusual for a trial involving surgical procedures. As all surgeons were able to carry out ThuVARP and TURP, the surgeon was blinded to the randomised allocation right up until the time of surgery. Following surgery, participants were not informed of their allocation but they were aware that this information was available from their general practitioner (GP) should they want to request it. Participants were asked at their 12-month follow-up which treatment they thought they had received and whether or not they had become unblinded at any point during the trial; these data were explored and utilised in a sensitivity analysis.

All investigators remained blinded to aggregate data throughout the recruitment and analysis of participants. The senior statistician, Chris Metcalfe, had not seen any data while writing the statistical analysis plan and remained blinded until the analysis had been finalised. The junior statistician (GY) had access to a random small subset of participants (n = 20) while writing the analysis plan to allow her to become familiar with the question and answer layouts. The protocol was written before recruitment ended and was published in Trials.1 The statistical analysis plan18 was written and agreed by the trial team in October 2017, post recruitment end but prior to any statistical analysis. The junior statistician became unblinded with full access to the data on commencement of analysis in January 2018.

Data collection

Case report forms (CRFs) were collected at baseline, perioperatively, postoperatively, 3 months after surgery and 12 months after surgery (Table 1). Questionnaires were given to participants to fill in at their baseline clinic, by post at 6 weeks, and at their 3- and 12-month clinics.

Participants with indwelling catheters were unable to complete urine flow tests; participants using intermittent catheterisation provided flow tests if they were able to void.

The trial team became aware that quite a few CRFs and questionnaires were being filled in much earlier or later than their scheduled follow-up. Therefore, before analysis the team established a time frame for which CRFs/questionnaires would be included in the analysis, as detailed in Table 2.

Patient and public involvement

We involved patient and public involvement representatives at all stages of the project. We discussed the design of the trial with several men awaiting prostate surgery for BPO before we submitted the grant proposal. Discussion subjects included the comparative nature of the trial and the acceptability of randomisation, and the burden of the questionnaires.

Throughout the trial, two patient representatives were members of the UNBLOCS Trial Steering Committee, and they attended all meetings. They advised on recruitment and retention, provided feedback on patient-facing materials and newsletters and were directly involved in creating a press release to raise the profile of the trial.

Statistical methods

The main statistical analyses were prespecified using a statistical analysis plan.18 Final data analysis started in January 2018 and finished in March 2018. Stata version 15.1 (StataCorp LP, College Station, TX, USA) was used for all statistical and health economic analyses in this trial. Binary outcomes were presented as numbers and percentages, whereas continuous outcomes were presented as mean and SD or median and interquartile range (IQR), as appropriate.

Primary analysis

Maximum urine flow rate was collected in clinic approximately 12 months after randomisation, while the IPSS was patient reported and collected using questionnaires sent to the patient approximately 12 months after randomisation. The primary analyses were conducted under the intention-to-treat principle using multivariable linear regression and adjusting for the stratification variables used in randomisation (centre and baseline LUTS/retention). The team originally planned to impute missing 12-month IPSS and Qmax data using data from 6 weeks and 3 months; however, the statistical nature of these data meant that this was not possible (see Multiple imputation). Instead, a conditional multiple imputation model using chained equations was employed and used data from baseline, including all relevant IPSS data at baseline, baseline flow, comorbidities, age, baseline diagnosis and arm. At baseline and 12 months, conditional imputation allowed the model to impute only for participants with intermittent catheters or no catheters. Participants unable to void with indwelling catheters were unable to complete urine flow tests and the IPSS questions were not relevant.

The results were based on the equivalence margin prespecified in the trial design process. As the primary analysis tested equivalence, interpretation of primary analysis results focused on observed differences and 95% CIs for the between-group comparisons. When CIs lay within the equivalence margins, the two arms were deemed equivalent. When equivalence was not achieved, the team tested for superiority, as this does not incur a statistical penalty.23

• H₀: For IPSS and Qmax, ThuVARP is better or worse than TURP by more than 2.5 points or 4 ml/second, respectively.
• H₁: For IPSS and Qmax, ThuVARP is not better or worse than TURP by more than 2.5 points or 4 ml/second, respectively.

Secondary outcomes

All secondary analyses, unlike the primary analyses, tested for superiority. Therefore, p-values were presented alongside observed differences and CIs. All analyses were adjusted for centre and diagnosis at baseline (LUTS/retention), apart from the Clavien–Dindo19 grading of complications (based on the small number of events).

• Prespecified surgical complications were collected in CRFs both at the perioperative stage and at each time point when a clinic CRF was administered. The Clavien–Dindo (1–5) classification scale of surgical complications was used to grade the severity of each complication. The number of complications experienced per participant was explored, along with the worst event per participant, using ordinal logistic regression. Given that complications were repeatedly reported across CRFs, the data were not explored at the event level. The complications were graded using the following system:
  1. deviation from normal postoperative course without the need for further interventions (pharmaceutical, surgical, etc.)
  2. requiring pharmacological treatment
  3. requiring surgical intervention (a) not under general anaesthetic or (b) under general anaesthetic
  4. life-threatening complication: (a) single-organ or (b) multiorgan failure
  5. death of the participant.
• The number of hours that the participant remained in hospital after surgery and the drop in blood haemoglobin/sodium levels were compared between the two arms using linear regression. When the distributions of continuous variables were skewed, medians and IQRs were presented alongside regression models (after checking residuals). The p-value was also compared with the one achieved from the Mann–Whitney U-test to ensure consistency. The number of hours until successful trial without catheter was analysed using a Cox proportional hazards model, whereby the assumption of proportional hazards was checked using a Schoenfeld residuals test.24 The requirement of a blood transfusion was recorded perioperatively, postoperatively and 3 months and 12 months after surgery. The proportion of participants requiring a transfusion within 12 months of surgery was compared between the arms using logistic regression. Catheter use was recorded at 3 months and 12 months and the proportion of men using a catheter was compared using logistic regression, for both time points. We intended to compare post-void residual at 12 months using linear regression; however, after inspecting the results, the decision was made to compare the groups using logistic regression (zero vs. non-zero residuals) and ordinal logistic regression, as a large proportion of participants had post-void residuals of zero.
• Each of the following patient-reported outcome measures was compared between the arms at 12 months using linear and logistic regression as appropriate. When the distributions of continuous variables were skewed, medians and IQRs were presented alongside regression models (after checking residuals). The p-value was also compared with that achieved from the Mann–Whitney U-test to ensure consistency:
  • IPSS quality of life
  • ICIQ-MLUTS
  • ICIQ-MLUTSsex
  • ICIQ-LUTSqol
  • ICIQ-satisfaction
  • IIEF-5.

A voiding and incontinence score can be created from the ICIQ-MLUTS questionnaire and an overall erectile dysfunction score can be created from the IIEF-5. Similar to previous studies, the following dichotomous variables were created for ease of reporting25 (Table 3). The results were also compared on an ordinal scale to ensure that they remained the same.

Sensitivity analyses

Several sensitivity analyses were conducted to test the robustness of the results to different assumptions made.

• Complete-case analysis: the primary analysis was repeated without imputation for missing variables. This set was then used to compute the following sensitivity analyses.
• Per-protocol analysis: using complete cases, the per-protocol analysis included only those patients who received their randomised treatment. Those who had converted from ThuVARP to TURP were not included.
• Complier average causal effect (CACE) analysis: using complete cases, the CACE analysis utilised all data but, unlike the primary analysis, incorporated the treatment received variable as the independent variable and randomisation as an instrumental variable using the ‘iv regress 2sls’ command in Stata.
• Removal of participants: nine participants found out their allocation at their 12-month appointment (not including those patients who guessed correctly) before they completed the 12-month questionnaire. As the site confirmed that these participants had become unblinded, they were removed in a sensitivity analysis.
• Adjustment for baseline: as baseline IPSS and Qmax were not available for those patients who had an indwelling catheter, the primary analysis did not adjust for these baseline measures to avoid a large number of missing or imputed data. In a sensitivity analysis, participants with an indwelling catheter at baseline were given an imputed Qmax value of zero. For IPSS, there was no specific single value that made sense for catheterised participants. Instead, the baseline scores were categorised, in accordance with the American Urological Association classification,26 into mild (0–7), moderate (8–19) or severe (20–35). Those participants with an indwelling catheter were placed in the ‘severe’ category.
• Adjustment for imbalance at baseline: covariates that differed at baseline by more than half a SD (or by 10%) were adjusted for in an additional model to investigate their effect on the difference observed between the two groups.
• Type of TURP/surgery: ThuVARP was compared separately with monopolar and with bipolar TURP. The two TURP procedures were also compared in terms of IPSS and Qmax, and therefore as three two-way tests. However, the results were interpreted with caution as this analysis was exploratory.
• Surgeon effects: a mixed-effects model was conducted that included the surgeon as a random effect and the centre as a fixed effect in the primary analyses.
• As a post hoc sensitivity analysis, bootstrap regression (4999 replications) was performed to allow for the slightly skewed distribution of IPSS and Qmax.
• As a post hoc sensitivity analysis, given the slightly skewed distribution of IPSS and Qmax, log-transformations were carried out. For the IPSS, 1 was added to all scores before being log-transformed (natural log). Qmax scores were directly log-transformed (natural log). The equivalence margins were also log-transformed to aid interpretation of CIs, 0.92 and 1.39 for IPSS and Qmax, respectively.

Subgroup analyses

Prespecified subgroups were used to test whether or not the difference between ThuVARP and TURP was more pronounced in certain subgroups of patients. Although underpowered, tests of interaction between the dichotomised/categorical variables and treatment were carried out to explore whether or not treatment effect differed between subgroups; the focus of interpretation is on the CIs rather than the p-values. These interaction terms were added to the primary analysis model.

Subgroup analyses included:

• baseline diagnosis of LUTS or retention
• age (split by median age)
• preoperative prostate size in ml, estimated during digital rectal examination (small, < 40 ml; medium, 40–60 ml; large, 60–80 ml; and very large, > 80 ml)
• patients with or without comorbidities at baseline.

In the protocol we had specified that we would look at the length of stay of procedures (day case or inpatient). However, it was later decided that this would be more suitable as an outcome only, as the baseline intention would be unlikely to alter the treatment effect on IPSS/Qmax.

Multiple imputation

Multiple imputation by chained equations was used to impute missing values for the primary outcomes. Originally, as prespecified in the analysis plan, the model intended to use data collected at 6 weeks and 3 months to inform the 12-month data. However, when these data were included, the model failed to converge owing to the high collinearity between the measures. On inspection of the CONSORT (Consolidated Standards of Reporting Trials) flow chart (see Figure 1), the team felt that this was not detrimental to the study as few participants had completed 3-month questionnaires but not 12-month questionnaires. Instead, the model included baseline and 12-month follow-up variables to inform imputation. Trial arm, baseline diagnosis of LUTS or urinary retention, baseline comorbidities and age were included as complete variables. Indwelling catheter status (yes/no) was imputed at baseline and 12 months, while Qmax and IPSS individual items at baseline and 12 months were imputed conditionally when no indwelling catheter was present at the corresponding time point. The general rule of thumb is that the number of imputations should exceed the percentage of incomplete cases; therefore, for this trial, 40 individual imputations were created and combined using Rubin’s rules in Stata 15.1.27 The prespecified randomisation seed 525 was used to create reproducible imputations.