Effectiveness of Community-based Football in Prostate Cancer

PLAN FOR THE STATISTICAL ANALYSES

This plan has been prepared and is laid forward before any follow-ups are conducted. The plan delineates the analyses of the FC PC trial; any deviations will be laid forward and discussed in the final publication.

The primary statistical analysis targets the effect on prostate cancer (PCa) specific Quality of Life (QoL) of a treatment policy offering community based football to men with prostate cancer. In particular, patients will be analysed in the treatment group to which they were randomly allocated respecting the intention-to-treat principle (ITT). The appropriate method for addressing this effect depends on the assumptions made about missing data (drop-outs). Our main analysis is valid under the missing at random assumption but we will also present sensitivity analyses robust to non-ignorable patterns among patients with incomplete data.

Per protocol analyses will be performed to estimate the de jure effect of the treatment for compliant patients. The "per protocol" population will be defined as those trial participants from the intervention group that have attended the football intervention at least 12 times in the first 12 weeks and 24 times in 6 months intervention period.

Significance tests will be two-sided with a maximal type I error risk of 5 %. To address the problem of multiple comparisons for secondary analyses when several outcomes are tested or multiple constracts are extracted from the same statistical model p-values will be adjusted using the step-down Bonferrroni method of Holm (Holm 1979) or appropriate modern alternatives.

Trial profile A CONSORT diagram will show trial participant flow. Number of screened patients fulfilling inclusion criteria and trial subjects included in analyses together with reasons for exclusion of trial subjects will be reported.

Primary outcome The continuous outcome score of FACT-P will be calculated using the official scoring guideline. As described in the scoring guideline missing items will be prorated by multiplying the sum of the subscale with the number of the items in the subscale, then divided by the number of items answered. This will only be done if more than 50% of the items are answered in the subscales and 80% are answered in the total questionnaire. The change score of the total FACT-P at 12 weeks will be calculated by subtracting the total 12 week score from the respective trial participant's baseline score. Analysis of covariance will be used (Vickers and Altman 2001), group and ADT-status will be set as factors, the response will be change in FACT-P and covariates are age and baseline score. The results will be presented as least squares means (LSMEANS) differences between the two groups with 95 % confidence intervals and p-values.

Secondary outcomes including safety outcomes Changes from baseline to 6 month for LBM, BMC, BMD, and total body fat mass will be analysed in the same manner as the primary outcome.

QoL, Functional Well-Being and Physical Activity (based on MET values derived from the IPAQ questionnaire) measured at baseline, 12 weeks and 6 months will be analysed using a mixed model for repeated measurements. Changes from baseline to 12 weeks or 6 months will be treated as the response, the model will include fixed effects of factors: group, ADT, sampling time and their interactions, and the analysis will be adjusted for age and baseline value. The correlation between measurements on the same patient will be modeled using a random effect of patient.

Safety outcomes will be listed for each group and the number of fractures and falls that resulted in seeking medical assessment will be compared across groups using Fisher's exact test.

Subgroup analyses are planned to be reported for the patients treated with androgen deprivation therapy therefore results will be given both for the overall treatment effect and for subgroup obtained by stratifying level of according to androgen deprivation therapy. To verify the credibility of our subgroup analyses, we apply the criteria proposed by Sun and colleagues (Sun et al. 2012), i.e. the subgroup variable is a baseline characteristic, is a stratification factor, is specified a priori and includes only a small number of analyses.

Outline of figures and tables The first figure in the main publication will be a CONSORT flow diagram. The second figure will illustrate changes in the primary and secondary outcomes, except safety outcomes, at 12 weeks and 6 months, according to treatment group.

A third figure will display mean curves for the primary outcome for patients in different groups according to pattern of missing data. In particular, mean curves will be shown separately for completers and patients with missing data at one or more assessment times. The figure will be used to guide the type of sensitivity analyses performed to adjust results for a potentially deviating pattern for patients with incomplete observations.

Tables will be in following order; first a table delineating trial subjects characteristics, secondly a table showing changes in primary and secondary outcomes both at 12 weeks and 6 months. Thirdly safety outcomes will presented according to group and type.