The randomised thoracoscopic talc poudrage+indwelling pleural catheters versus thoracoscopic talc poudrage only in malignant pleural effusion trial (TACTIC): study protocol for a randomised controlled trial

Introduction Malignant pleural effusion (MPE) is common, with 50 000 new cases per year in the UK. MPE causes disabling breathlessness and indicates advanced disease with a poor prognosis. Treatment approaches focus on symptom relief and optimising quality of life (QoL). Patients who newly present with MPE commonly require procedural intervention for both diagnosis and therapeutic benefit. Thoracoscopic pleural biopsies are highly sensitive in diagnosing pleural malignancy. Talc poudrage may be delivered at thoracoscopy (TTP) to prevent effusion recurrence by effecting pleurodesis. Indwelling pleural catheters (IPCs) offer an alternative strategy for fluid control, enabling outpatient management and are often used as ‘rescue’ therapy following pleurodesis failure or in cases of ‘trapped lung’. It is unknown whether combining a TTP with IPC insertion will improve patient symptoms or reduce time spent in the hospital. The randomised thoracoscopic talc poudrage + indwelling pleural catheters versus thoracoscopic talc poudrage only in malignant pleural effusion trial (TACTIC) is the first randomised controlled trial (RCT) to examine the benefit of a combined TTP and IPC procedure, evaluating cost-effectiveness and patient-centred outcomes such as symptoms and QoL. The study remains in active recruitment and has the potential to radically transform the pathway for all patients presenting with MPE. Methods and analysis TACTIC is an unblinded, multicentre, RCT comparing the combination of TTP with an IPC to TTP alone. Co-primary outcomes are time spent in the hospital and mean breathlessness score over 4 weeks postprocedure. The study will recruit 124 patients and aims to define the optimal pathway for patients presenting with symptomatic MPE. Ethics and dissemination TACTIC is sponsored by North Bristol NHS Trust and has been granted ethical approval by the London-Brent Research Ethics Committee (REC ref: 21/LO/0495). Publication of results in a peer-reviewed journal and conference presentations are anticipated. Trial registration ISRCTN 11058680.


Screening
Baseline assessment  The primary analysis for each outcome will be by intention-to-treat, meaning that all patients on whom an outcome is available will be included in the analysis, and will be analysed according to the treatment group to which they were randomised. All tests will be two-sided and will be considered statistically significant at the 5% level.
For each analysis, the following summaries will be provided: a. The number of patients in each treatment group who are included in the analysis. The mean (SD) or median (IQR) in each treatment group for continuous outcomes, or the number and percentage of patients experiencing an event for categorical or time-to-event outcomes (time to-event outcomes will also present the median time to event in each treatment arm if applicable).
b The treatment effect (difference in means for continuous outcomes, standardized effect size, odds ratio for binary outcomes, hazard ratio for time-to-event outcomes, rate ratio for count outcomes) with its 95% confidence interval and a p-value.
The primary outcomes will be initially compared between groups (raw analysis) on an intention to treat basis. Adjusted analysis using regression for baseline imbalance and the minimisation factors will then be conducted to increase statistical power and certainty. Adjusted analyses for the minimisation variables will include the minimisation variables as covariates in the relevant regression model and will include i) type of underlying malignant disease [mesothelioma, other, unknown] and ii) WHO performance status [0-1 or 2-3]).
Missing data will be minimised but major remaining missing data (e.g. primary outcome) will be dealt with using sensitivity analyses.

Interim analysis
No Interim analyses are planned for this study. In the absence of a Data Monitoring Committee, the Oxford Respiratory Trials Unit (ORTU) Safety Oversight Group will advise the TSC on issues related to any safety data arising from the trial.

Sample size calculation
The study has two co-primary outcomes and separate sample size calculations are presented below. The null hypothesis states that there is no difference between the intervention (TTP+IPC) and standard care (TTP alone) in length of hospital stay over the first 4 weeks post randomisation, and in patient reported breathlessness over the first 4 weeks post randomisation.

Hospital stay:
Previous randomised studies demonstrate mean initial hospital stay in IPC treated patients of 0 days (SD 1) and in talc slurry pleurodesis of 4 days (SD 2). As talc poudrage pleurodesis is likely to result in shorter hospital stay than slurry, and this study captures hospital stay over 4 weeks rather than initial stay only, we have here conservatively assumed mean TTP+IPC stay over 4 weeks of 1 day, and mean TTP stay of 3 days, with a shared SD of 3 days to capture variability. Using these assumptions, a total of 124 participants are required (two-sided alpha=5%, 95% power, including 5% attrition due to nonmalignant diagnosis). This detected difference encompasses a clinically significant effect which would influence clinical decision making (i.e. reduction in hospital stay of 2 days total over one month, identified in patient surveys as a priority).

Breathlessness:
Previous randomised MPE studies demonstrate the mean breathlessness score in patients post treatment over 6 weeks is 25mm (SD 26mm) when treated with either talc or IPC. The MCID for VASd is 19mm. To detect a smaller difference than the MCID (to be conservative) of 16mm requires a total of 116 participants (two-sided alpha=5%, 90% power, including 5% attrition as above). Thus, the planned sample size of 124 participants is over-powered for this outcome.

Total number of days spent in hospital over 4 weeks post treatment.
(a) Unadjusted analyses will compare total number of days spent in hospital 4 weeks post treatment between randomised arms using the non-parametric log-rank test accounting for any censored data due to death defaulting to the Mann Whitney test in the absence of censoring. 95% confidence intervals for median differences will be reported. (b) Negative binomial regression will be used to compare randomised groups on this outcome including the minimisation factors as covariates.

Average breathlessness over 4 weeks post treatment
(a) Unadjusted analyses will compare average breathlessness over 4 weeks post treatment between randomised arms using the non-parametric Hodges-Lehman test. (b) Linear regression (ANCOVA) will be used to compare randomised groups on this outcome including the minimisation factors as covariates. (c) Adjusted analyses will compare average breathlessness over four-week period between randomised arms using ANCOVA with baseline breathlessness and minimisation factors as covariates.
Conclusions will be drawn from the adjusted analyses. (a) VAS breathlessness immediately post procedure will be compared between randomised arms using ANCOVA with baseline VAS breathlessness as a covariate. 1.5.2. Time spent in hospital over 12-weeks.
(a) Unadjusted analyses will compare total number of days spent in hospital over 12 weeks between randomised arms using the non-parametric log-rank test accounting for any censored data due to death defaulting to the Mann Whitney test in the absence of censoring. 95% confidence intervals for median differences will be reported. (b) Negative binomial regression will be used to compare randomised groups on total number of days spent in hospital over 12 weeks including the minimisation factors as covariates.

Time in hospital over 4 weeks accounting for participants who are suitable for discharge but remain in hospital for non-medical reasons.
Time spent in hospital not counting time in hospital when otherwise suitable for discharge will be analysed. (a) Unadjusted analyses will compare total number of days spent in hospital 4 weeks post treatment between randomised arms using the non-parametric log-rank test accounting for any censored data due to death defaulting to the Mann Whitney test in the absence of censoring. 95% confidence intervals for median differences will be reported. (b) Negative binomial regression will be used to compare randomised groups on this outcome including the minimisation factors as covariates.

Time in hospital over 12 weeks accounting for participants who are suitable for discharge but remain in hospital for non-medical reasons.
Time spent in hospital not counting time in hospital when otherwise suitable for discharge will be analysed. (a) Unadjusted analyses will compare total number of days spent in hospital 4 weeks post treatment between randomised arms using the non-parametric log-rank test accounting for any censored data due to death defaulting to the Mann Whitney test in the absence of censoring. 95% confidence intervals for median differences will be reported. (b) Negative binomial regression will be used to compare randomised groups on this outcome including the minimisation factors as covariates.

Average Chest Pain over four weeks post treatment.
(a) Unadjusted analyses will compare average VAS Chest Pain over post 4 weeks post treatment between randomised arms using the non-parametric Mann Whitney test with 95% Hodges-Lehman confidence intervals. (b) Adjusted analyses will compare VAS Chest Pain immediate post procedure between randomised arms using ANCOVA with baseline VAS Chest Pain as a covariate. (c) Adjusted analyses will compare VAS Chest Pain immediate post procedure between randomised arms using ANCOVA with baseline VAS Chest Pain and minimisation factors as covariates. 1.5.6. Chest Pain Immediately Post Procedure.
(a) VAS Chest Pain immediate post procedure will be compared between randomised arms using ANCOVA with baseline VAS breathlessness as a covariate.

Pleurodesis success at 4 and 12 weeks.
(a) Unadjusted analyses will compare proportion and percentage of those achieving successful pleurodesis at 4-weeks between randomised arms using Fisher's exact test. 95% confidence intervals for the difference in success rates will be constructed using the Miettinen-Nurminen formula. 95% confidence interval for the proportion/percentage of successful pleurodesis at 4 weeks within each arm will be constructed using the Wilson score approach.
The same analyses will be conducted for successful pleurodesis at 12-weeks. (b) Adjusted analyses will compare proportion of those achieving successful pleurodesis at 4weeks between randomised arms using binary logistic regression with minimisation variables as covariates. The odds ratio and 95% confidence interval will be reported. The same analysis will be conducted at 12-weeks. 1.5.8. Number of healthcare contacts post discharge over 12 weeks.
(a) Unadjusted analyses will compare number of primary care visits at 12-weeks post procedure between randomised arms using the non-parametric Hodges-Lehman test. (b) Negative binomial regression will be used to compare randomised groups on this outcome including the minimisation factors as covariates. 1.5.9. Quality of life at 4-and 12-weeks post procedure.
EQ-5D-5L will be scored to give quality of life indices and EQ-5D-5L VAS rating. The EORTC QLQ-30 comprises 30 items, 24 of which are aggregated into nine multi-item scales, comprising five functioning scales (physical, role, cognitive, emotional and social), three symptom scales (fatigue, pain and nausea/vomiting) and one global health status scale. The remaining six single-item (dyspnoea, appetite loss, sleep disturbance, constipation, diarrhoea and the financial impact) scales assess symptoms. Scales and subscales are scored 0 -100. All scales and subscales will be analysed in the same way at both 4-and 12-weeks. (a) Unadjusted analyses will compare QoL measures between randomised arms using the nonparametric Hodges-Lehman test. (b) Linear regression will be used to compare randomised groups on QoL measures including the baseline measure as a covariate. (c) Linear regression will be used to compare randomised groups on QoL measures including the minimisation factors and baseline measure as covariates.  Burden is assessed using the Physical Functions,Physical Limitations,Bodily Pain,General Health,Vitality,Social Functioning,Emotional problems,Mental Health each transformed between 0 -100) and the General Health Questionnaire. All measures will be analysed in the same way at each of 4-weeks and 12-weeks. (a) Unadjusted analyses will compare average burden at 4-weeks post treatment between randomised arms using the non-parametric Hodges-Lehman test. This analysis will be repeated at 12-weeks. (b) Linear regression will be used to compare randomised groups on each measure of burden including the minimisation factors as covariates.

Subgroup analyses
Analyses will be conducted on the Intention-to-Treat analysis set. In addition, two pre-specified subgroups analyses will be undertaken. One subgroup will be all patients excluding those with trapped lung. A second subgroup will be all patients excluding those with a prolonged air leak after thoracoscopy -these patients could skew results. These subgroup analyses would be on the coprimary outcome measures. Any other subgroup analysis would be exploratory.

Sensitivity analyses
Primary analyses will be conducted using the Intention-to-Treat analysis set. The sensitivity of statistical conclusions with respect to the inclusion/exclusion of any clinically unusual, interesting, challenging or outlying cases will be undertaken including death. The robustness of statistical conclusions with respect to missing data will but undertaken on a missing-not-at-random (MNAR) basis using best-worst worst-best scenarios. If <5% of data is missing, then complete/available case analyses will be conducted. If 5% to 30% of data is missing, then best-worst worst-best sensitivity analyses will be undertaken. If more than 30% of data is missing, then results will be considered as hypothesis generating.