Specialised palliative care (SPC) seeks to relieve suffering and improve quality of life in patients with a life threatening disease such as advanced cancer. Many patients with advanced cancer are not in contact with SPC. Previous studies have shown that among advanced cancer patients not referred to SPC there is a significant prevalence of symptoms, problems and needs. The aims of the present study are to investigate whether patients with metastatic cancer, who report palliative needs in a screening, will benefit from being referred to SPC and to investigate the economical consequences of such a referral.

The trial is a randomised, clinical, multicenter trial including 6 Danish SPC-centres. The basic principle is that patients with palliative needs (see inclusion criteria) are identified at oncological departments and randomised to either (i) standard treatment plus SPC (intervention group) or (ii) standard treatment (control group).

Patients will be identified by the following procedure: A) Each week a research nurse reviews the medical records of consecutive patients seen in oncological out-patients clinic. B) Eligible patients are asked to fill out a questionnaire (the screening) that investigates symptoms and problems. The patients are told that their questionnaire will be evaluated and that the research nurse will contact some of the patients later on with information about a RCT. C) Those patients who report at least one palliative need in the questionnaire (see inclusion criteria) and have four additional symptoms are contacted by the research nurse who provide the patients with written and verbal information about the RCT. D) Patients who give informed consent are randomised.

Detailed statistical analysis plan:

Analysis of the primary outcome The primary outcome analysis will be a modified ITT analysis. Patients who withdrew consent after randomisation, who were randomised by mistake and did not fulfil our inclusion criteria, or who were not alive at the time of the follow-ups, will be excluded from the primary analysis. All exclusions will be shown in the CONSORT flow-chart of patient participation.

In the primary outcome analysis we will use multiple imputation for non-responders if there are more than 5% missing outcomes. In total, we will make 20 different data-sets with imputation based on a regression model using predictive-mean-matching using the MI and the MI ANALYSIS procedures in SAS. The primary outcome analysis will be made as a multiple regressions adjusted for the stratification variable if it is normally distributed.

Sensitivity analyses of the primary outcome We will make five sensitivity analyses: 1) a fully adjusted analysis including all relevant covariates, 2) a complete case analysis, 3) an analysis including a model for repeated measurement, 4) a per protocol analysis, 5) an analysis including imputations for those who died.

Analysis of secondary outcomes The analyses of the seven scales from EORTC QLQ-C30 (physical function, role function, emotional function, nausea and vomiting, pain, dyspnoea, or lack of appetite) will be made using the same principles as described for the primary outcome including the sensitivity analyses. Survival will be analysed using Kaplan-Meier plot. Patients who are alive three months after the end of data-collection will be censured on this date. A Cox regression will be made adjusted for the stratification variable. A sensitivity analysis will be made adjusting for all relevant covariates.

Exploratory outcomes and subgroup analyses For serious adverse events we report the number of hospitalisations, the number of acute hospitalisations and the number of deaths in the eight week trial period. The analyses of the other exploratory outcomes will not be dealt with in detail here. The overall principles regarding questionnaire data (the remaining scales from EORTC QLQ-C30, HADS and FAMCARE-P16) are that they will be analysed as complete case analyses.

Significance levels All tests will be two-tailed. For the primary outcome the risk of type I error is set to 5% (i.e., a significance level of P<0.05). As we have 8 secondary outcomes, we adjust the significance levels to P<0.01 to control the familywise (or cumulative) type I error due to multiplicity. The P-values of the exploratory outcomes will be provided, but it will be made clear that the analyses are exploratory.

Register-data, data-management and analyses Survival will be retrieved from the Danish Civil Registration System (CPR), and serious adverse events and contacts from The Danish Patient Registry (Landspatientregistret).

Data management will be done by project manager Anna Thit Johnsen. Analyses will be made by statistician Morten Aa. Petersen, who is blinded to the identity of the two intervention groups, which will be denoted Y and X. Results will be presented blinded in the same way for the investigators, and conclusions regarding the results will be drawn by the investigators and written down while the interventions are still blinded. The blinding will not be broken before all analyses of primary and secondary outcomes have been conducted.

Inclusion Criteria:

• Cancer stage 4 according to the TNM system (for patients with cancer in CNS inclusion criteria is: cancer grade three or four and no possibility of radical treatment)
• At least 18 years
• Live in the area of the participating hospitals
• No contact with specialised palliative care within the previous year
• At least one palliative need defined as a EORTC QLQ-C30 scale score of at least 50% of the score corresponding to maximal symptomatology or maximal functional impairment
• Four additional symptoms (defined as patients answering, on average, at least 'a little' in any of the 14 function or symptom scales in the EORTC QLQ-C30 questionnaire)
• Written informed consent

Exclusion Criteria:

• Do not understand Danish well enough to participate in the study
• Are judged incapable of co-operating with the trial protocol