Endoscopic sinus surgery in adult patients with Chronic Rhinosinusitis with nasal polyps (PolypESS)-statistical analysis plan for a multicentre randomised controlled trial*

Background: Chronic rhinosinusitis with nasal polyps (CRSwNP) afflicts 2-4% of the population and comes with a long time burden of disease and high societal costs. The current treatment consists of medical treatment alone or in combination with endoscopic sinus surgery. No consensus exists on the right timing and extent of disease that warrants surgery. Furthermore, there is lack of clinical knowledge about the benefit of surgery over medication only. The current study evaluates the clinical effectiveness and cost-effectiveness of endoscopic sinus surgery in addition to medical treatment versus medication alone in the adult patient group with nasal polyps (CRSwNP). Methods: The PolypESS trial is designed as a prospective, randomised, multicentre trial in adult patients with CRSwNP selected for primary or revision endoscopic sinus surgery by their otorhinolaryngologist. Patients are randomly assigned to endoscopic sinus surgery in addition to medication or medical therapy only. This paper details the statistical analysis plan (SAP) of this trial and was submitted before outcome data were available. Results: The primary outcome of the trial is disease-specific Health-Related Quality of Life quantified by the SNOT-22 at 12-months follow-up. Secondary outcomes consist of generic and disease-specific Health-Related Quality of Life, objective signs of disease and adverse events of treatment. Subgroup analyses will be performed to verify if treatment effects differ among patient phenotypes. Analyses will be completed according to this pre-specified SAP. The main analysis will be performed as a standard ITT analysis. Discussion: The PolypESS trial will show whether addition of endoscopic sinus surgery to medical treatment improves the disease-specific Health-Related Quality of Life quantified by the SNOT-22 at 12-months follow-up. Unforeseen deviations from the SAP at the time of analysis will be motivated and discussed in the final publication of the primary outcome of this study.


Introduction
Patients with chronic rhinosinusitis with nasal polyps (CRSwNP) experience a significant impact on most aspects of their Healthrelated Quality of Life (HrQOL) (1)(2)(3) . Together with the high prevalence of disease (1-4%), the diagnosis and treatment give high healthcare costs (4)(5)(6) . The current treatment exists of medical treatment alone or in combination with endoscopic sinus surgery (ESS) (5,(7)(8)(9)(10)(11) . It is unclear what the benefit is of the addition of surgical treatment on top of on-going medical treatment and at what time point in the disease course a patient should be offered surgery. The PolypESS study is the first multicentre, randomised controlled trial investigating the impact of ESS on disease-specific HrQOL in adult patients suffering from CRSwNP in comparison to on-going medical treatment. Currently a large RCT comparing ESS with a prolonged course of Claritromycine in patients with chronic rhinisinusitis without nasal polyps (CRSsNP) and CRSwNP is conducted in the UK (12) . Further details on the background of our current study are described in the previously published trial protocol (13) .

Study objectives
The primary objective is to assess the effect of performing ESS in addition to medical treatment instead of medical treatment alone on patient health-related quality of life (HrQOL) and cost-effectiveness in adults with CRSwNP. Primary hypothesis is that the addition of ESS is better than medical treatment alone considering the mean difference (95% CI) in total SNOT-22 score at 12 months follow-up. We will test for superiority. The secondary hypotheses will be evaluated for risk difference (%) or mean difference (95% CI) between intervention groups. The following secondary hypotheses will be tested: ESS is better than medication only in improving generic HrQOL (as measured with the EQ-5D-5L), ESS is better in improving objective signs of disease  (14) ), better asthma control (as measured with the Asthma Control Test (15) ) and less symptomatic exacerbations requiring further treatment including ESS at 12 months follow-up. We will descriptively report (serious) adverse events in both treatment groups. We hypothesize more adverse events in the medical treatment group at 12 months follow-up. For more details on the process of data collection and a description of all secondary outcome measurements we refer to the published study protocol article (13) .

Protocol developments
PolypESS is an investigator-initiated, prospective, open, multicentre randomised clinical superiority trial with parallel treatment groups. Participants are randomised to either ESS in addition to medical treatment or medical treatment alone.
Medical treatment can be any treatment available for CRSwNP.

General principles
The analyses will be done by the investigators of the study group supervised by an independent statistician. The analyses will be performed after data verification and validation have been carried out and after this SAP has been accepted for publication. The statistical programming and analysis to produce all tables and figures will use the SPSS v. 26 (IBM Corporation, Armonk, NY, USA) and the software environment R (latest version 4.0.3) (17) . Descriptive statistics, means with SD for continuous normally distributed variables, medians and interquartile ranges for continuous skewed variables, and frequency counts with percentages for nominal variables will be used to summarize variables. Normality will be checked for with a Normal Q-Q plot and histogram. No statistical normality tests will be performed.

Patient flow diagram
A flow diagram of study participants will be displayed in line with the Consolidation Standard of Reporting Trials (CONSORT) recommendations and finalized upon external peer review ( Figure 1) (18) .

Treatment according to protocol and withdrawal
Treatment was regarded to have proceeded according to the study protocol if a patient had surgery or a discussion about additional medical treatment within 6 weeks after inclusion. All patients that attended the baseline visit will be included in the ITT population. Primary outcome is measured after 12 months, planned 12 months after the start of the allocated intervention.
For all time points within or at 12 months of follow-up a window of 30 days before or after the scheduled time point is accepted.
The numbers of losses to follow-up (withdrawal from follow-up) and dropouts (withdrawal from intervention) will be summarized by study arm. A line-by-line listing of reasons for withdrawal or loss to follow-up will be presented in an Appendix. A patient is considered lost to follow-up if both a scheduled study visit or replacement telephone visit could not be performed at 6 months follow-up and at 12 months follow-up (after at least three phone calls, two e-mails, sending postal questionnaires and a letter). If patients miss the 12 months visit, multiple imputations will be conducted if needed. A study visit is set to be missing if no SNOT-22 is obtained and the patient could not be contacted for study-related questions.

Definition of intention-to-treat, per-protocol and as-treated sets
The main analysis will follow the intention-to-treat (ITT) principle with all patients analysed in their randomisation group, irrespective of protocol adherence. This includes patients that crossed over to the other study treatment group during the course of the study (only possible from medication to ESS).
Only patients with a protocol violation concerning eligibility are excluded from the ITT analysis. Protocol violation in eligibility refers to randomised patients who did not fulfil inclusion criteria or randomised patients who did meet an exclusion criterion.
Baseline characteristics will be evaluated for these patients and compared to the ITT population. In addition, a per-protocol and as-treated analysis will be performed. Baseline characteristics will be compared between ITT,PP and as-treated with adjustment for confounding in the ITT and as treated analysis. The per-protocol analysis will include patients that were included and treated according to the study protocol. This means that patients who crossed over to the ESS treatment group will be excluded. The as-treated analysis includes patients that switched treatment (from medical to surgical). A summary of the inclusion and exclusion of patients in the analysis sets is displayed in Figure 2.

Representativeness of study sample
The total number of participants that were eligible will be reported including distribution of gender, age and when available disease-specific health-related quality of life (SNOT-22). To evaluate whether the randomised group is representative for all eligible patients, a comparison will be made between patients who declined to participate but were willing to fill in a SNOT-22 questionnaire and the randomised population. Mean age, percentage of males and mean or median disease-specific healthrelated quality of life, measured at baseline, will be compared.

Sample size
The power analysis is based on the literature-based assumption that the minimal clinically important difference (MCID) for the SNOT-22 is 8.9 points (SD 20.0) (19) . A two-group t-test with a twosided p-value of 0.05, a power of 90% to detect a difference and an anticipated 10% loss of follow-up led to 238 patients needed for the main analysis.

Patient replacement and handling of missing data
Patients not fulfilling eligibility criteria resulting in the exclusion of the ITT analysis will not be replaced. An analysis of missing data will be performed to check for the assumptions regarding the missing data. In participants with missing data for the primary outcome (SNOT-22 at 12 months follow-up), multiple imputation will be used to predict the outcome if more than 60% of data is present (≤40% missing data). Considering the type of variables for which data could be missing and the nature of the trial, missing data will probably be missing at random and will be multiple imputed using chained equations (MICE).
Results for the primary outcome at 12 months will be compared to complete cases.

Baseline characteristics
The mock-up of the baseline characteristics table can be found in  medians and interquartile ranges will be used in case of nonnormal distributions. Mean SNOT-22 scores will be dealt with as described above. Other missing data will not be imputed. The number of patients in the variable row will be reported when more than five patients have missing data for the variable of interest. We will not test for differences between study groups.

Assessment and analysis of primary outcome
The mock-up of the analysis of primary and secondary outcomes is shown in Table 2. For the primary outcome, SNOT-22 at 12 months, first a descriptive analysis will be performed. The mean difference with 95% CI will be reported for each treatment group.
Analyses will be stratified by baseline nasal polyp size, CT-sinus Lund-Mackay score, presence or absence of NSAID-Exacerbated Respiratory Disease (N-ERD) and tertiary care centres versus secondary care centres. If potential modification of the effect of ESS is suspected, subgroup analysis will be done further by multiple regression.

Assessment and analysis of secondary outcomes
Following the strategy for the primary outcome as described above, secondary outcome measures will be analysed to further evaluate the added value of ESS over medication alone. These outcome measures are described in Table 2.

Analysis of safety outcomes
Safety outcomes are serious adverse events (SAE) and nonserious adverse events (AE). Both will be explored and reported for each treatment group, listed in a table, if they are related to study treatment or study activities.

Discussion
The aim of the PolypESS trial is to provide evidence regarding the effect of ESS in adult patients with CRSwNP. In this statistical analysis plan, we present the methods we will use to evaluate whether or not ESS is of additional value in the care of patients with CRSwNP. We have chosen the widely accepted SNOT-22 as primary outcome measure as it reflects our main interest: whether a patient reports a better HrQOL after surgery. In order to approach the real-life situation, patients from secondary and tertiary care hospitals are included whenever the treating otorhinolaryngologists would consider surgery to be indicated.
Following the real-life dogma in which patients may need addi- Table 1. Baseline characteristics (data will be reported for both groups, total N=238).  The definition of the SNOT-22 is presented in the published study protocol

Item Description Scale of measurement
The difference / contrast in absolute SNOT-22 score between treatment groups and accompanying 95% CI. In addition, mean delta SNOT-22 will be reported (change from baseline) Items on the SNOT-22 will be summed to calculate a total score (0-110). The delta will be calculated for each patient or for the treatment group depending on the amount of missing values.
Baseline and 12 months follow-up Analysis in ITT and PP analysis.
First a descriptive analysis will be performed. Mean difference with 95% CI will be reported.

Secondary outcomes
Generic HRQOL as measured with the EQ-5D-5L A questionnaire comprising five domains/questions: mobility, self-care, usual activities, pain or discomfort and anxiety or depression. The EQ-5D-5L can describe 3125 (55) unique health states. In addition, a VAS for health status is applied (0-100)

1.
Difference between treatment groups in utility scores 2.
Difference between treatment groups in VAS calculated from EQ-5D-5L A health state index score will be calculated from individual health profiles using the Dutch time trade-offbased health utility algorithm for the EQ-5D-5L.

months follow-up
Analysis in ITT.
Mean, standard deviation (SD) or median and interquartile range in case of skewed data will be provided for the study population by visit and by treatment. Items on the five questions will be summed to calculate a total score which represents a category of control Level of control: <20 = uncontrolled asthma, 20-24= controlled asthma, 25 = well controlled asthma 12 months follow-up Analysis in ITT. Percentage and frequency count will be reported.

Nasal polyp score
Left and right side of the nose is scored for size of nasal polyps (0-4 on both sides). For a description of the scoring system, see the published protocol.
Difference between treatment groups in percentage and count of each category.
Score of left and right side will be summed to get a total score.

months follow-up
Analysis in ITT. Percentage and frequency count will be reported.

Modified
Lund-Kennedy endoscopy score (MLK) Left and right side of the nose is scored for presence or absence of polyp, oedema and discharge (total score 0-12). For a description of the scoring system see, the published protocol.
Difference between treatment groups in mean total MLK score Scores for three items on each side of the nose will be summed to calculate a total score.

months follow-up
Analysis in ITT. Range, mean, SD or median and interquartile range in case of skewed data will be reported.

Modified
Lund Mackay Postoperative Endoscopy Score (MLMES) Left and right maxillary, ethmoid, sphenoid, frontal sinuses and olfactory fossa are scored for mucosal inflammation, mucus and purulent discharge (total score 0-100). For a description of the scoring system see, the published protocol.
Difference between treatment groups in mean total MLMES scores. Only for patients that underwent ESS in the past or as part of the study treatment.
Scores for five items on each side of the nose will be summed to calculate a total score.

months follow-up
Analysis in ITT. Range, mean, SD or median and interquartile range in case of skewed data will be reported.

Unit of measurement Calculations or transformations
Timing of measurement

Primary analysis
Nasal obstruction Peak nasal inspiratory flow method (PNIF) is used to quantify nasal obstruction. For a description of measurement, see the published protocol.
Difference between treatment groups in PNIF score No calculations needed. Only the highest value will be used for an individual patient.

months follow-up
Analysis in ITT. Mean, SD or median and interquartile range in case of skewed data will be reported.

Olfactory function
The 'Sniffin' Sticks Identification test is used to assess olfactory performance by a 12-odor identification test. For a description of measurement, see the published protocol.
Difference between treatment groups in percentage of normosmic, hyposmic and anosmic patients.
Correctly identified odours will be summed and classified as normosmic (11-12 correct), hyposmic (7-10 correct) or anosmic (0-6 correct) 12 months follow-up Analysis in ITT. Percentage and frequency count will be reported for the study population by visit and treatment. Adverse events will be summed between baseline and 12 months follow-up 12 months follow-up Analysis in ITT. Percentage and frequency count of adverse events will be reported for the study population by treatment. Number of people with an event will be reported in both treatment arms. In the Appendix a line listing will be added of all adverse events per treatment group.

Daily nasal symptoms
Nasal symptoms will be recorded by patients each day 2 weeks before a visit until 2 weeks after a visit (score 0-3 for headache/ facial pain, rhinorrhea, nasal congestion, loss of smell). For a description of scoring, see the published protocol.
Difference in weekly daily symptom scores between treatment groups The main daily symptom sum-score is calculated for each patient as the sum of all individual symptom scores, representing the sum of the severity of the most common nasal symptoms (0-12) 12 months follow-up Analysis in ITT.
Only patients with ≥ 4 observations per week will be included. Mean, SD or median and interquartile range in case of skewed data will be reported. tional treatment over time, the study protocol enables crossover from medical treatment alone to the addition of surgery. Still, we will analyse the data primarily in an intention-to-treat fashion as described here. Unforeseen deviations from the SAP at the time of analysis will be motivated and discussed in the paper describing the primary and secondary outcomes.

Authorship contribution
WJF is the principal investigator in the Academic Medical Centre and coordinating investigator with regard to all participating sites. ESL, SR and WJF prepared the manuscript. All authors read and approved the final manuscript.