Velikova 2004.

Methods	RCT; Unit of allocation: Patient; Stratified by: Site of cancer
Participants	Outpatients attending the oncology clinic to start cytotoxic or biologic treatment and expected to attend the clinic at least three times. Setting / country: Leeds Cancer Centre Medical Oncology Clinic at St James's Hospital ‐ Leeds / UK Type of cancer: Any type Phase of care: Treatment Sample size at randomisation: 286
Interventions	Regular quality‐of‐life (HRQL) assessments and feedback to physicians: touch‐screen questionnaires were completed by the patients in the waiting room before every medical encounter. The intervention questionnaires used were the European Organisation for Research and Treatment of Cancer‐Core Quality of Life Questionnaire, version 3.0 (EORTC QLQ‐C30), and the Hospital Anxiety and Depression Scale (HADS). Results were fed back to physicians in a graphic printout form. The physicians were trained in interpretation of the questionnaires. A manual was prepared, with description of scales, interpretation of scores, and explanation of the graphs. Structured meetings were conducted individually, with each physician to discuss the study and review examples of HRQL and clinical details of real patients. Posters with interpretative information were displayed in clinics. The physicians were asked to review and use the HRQL results during all intervention encounters, unless totally inappropriate. Control: (1) No touch‐screen measurement of HRQL before clinic encounters. (2) Attention‐control group: completion of HRQL questionnaires on touch‐screen computer, but no feedback to physicians.
Outcomes	Patient: QoL Process: Length of encounters with physician, medical decisions, non medical actions
Notes	Length of follow‐up: 6 months
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "The random assignment was unbalanced 2:1:1 in favour of the intervention group, and stratified by site of cancer in random permuted blocks (block size was 8). Random assignment was carried out by telephone, by the Administrative Office at Cancer Research UK Centre (Leeds)." Quote from author e‐mail message: "The randomisation was carried out using randomised permuted blocks, stratified by site of cancer. The randomisation list for each stratum was generated in advance by a statistician using STATA. At the time of randomisation for each patient, the oncologist telephoned the independent randomisation line and the patients were allocated to the next available treatment on the appropriate list."
Allocation concealment (selection bias)	Low risk	See quote first item.
Blinding (performance bias and detection bias) Functional status	High risk
Blinding (performance bias and detection bias) Physical status	High risk
Blinding (performance bias and detection bias) Psychological status	High risk
Blinding (performance bias and detection bias) Social needs	High risk
Blinding (performance bias and detection bias) Quality of life	High risk
Blinding (performance bias and detection bias) Use of services	Low risk
Incomplete outcome data (attrition bias) All outcomes	Low risk	Quote: "Characteristics of non respondents were compared with respondents using Chi‐2 and t tests." Comment: Proportions of patients that completed the 6 months study were similar between groups (Arm #1= 65%; #2 = 57%; #3=65%). Reasons for attrition were also similar (Fig. 2)
Selective reporting (reporting bias)	Low risk	All outcomes described in Methods are reported in Results.
Other bias	Low risk	No evidence of other bias.
Baseline outcomes similar?	Low risk	Quote: "The model included FACT‐G scores over time as the outcome variable; baseline FACT‐G score as a covariate; performance status, time, study arm, and study arm X time as fixed effects; and patient and patient X time as random effects." Comment: Table 1 demonstrate that baseline outcomes measures were similar between groups. For the FACT‐G measure, the time effect was evaluated so baseline difference were taken into account.
Baseline characteristics similar?	Low risk	Quote: "Table 1 presents the baseline patients and encounters characteristics, demonstrating a good balance of baseline variables between the study arms" Comment: Statistical tests were not performed to compare baseline characteristics between groups but results appeared fairly similar.
Protected against contamination?	High risk	Quote: "An optimal experimental design for the study was difficult to achieve for several reasons. The study was conducted in a natural environment (oncology clinics), with two groups of subjects (physicians and patients) who were in continuous complex interactions. The experimental intervention was both at patient level (completion of intervention questionnaires) and physician level (feeding back of HRQL information). Random assignment of physicians was considered, but rejected due to practical limitations. In the Cancer Centre, Leeds (similar to many large oncology practices in the United Kingdom), patient care is provided by teams consisting of four to seven physicians, and over time, patients usually see several different physicians who, if physicians were randomly assigned, might happen to be either in the experimental or the control group. If different clinics were randomly assigned instead of individual physicians, definite differences between patients would result, as the clinics were specialised by cancer site. Therefore, patients were chosen as the units of random assignment, with an analysis of possible physician‐sensitizing effect planned at the design stage."