Table 1 Characteristics of included studies

Angiolillo et al. (2004)[24] United States

Controlled observational study

Parents of children with cancer

Four Children Cancer Group Trials

I: n = 36;

C: n = 104

Parents of children with acute leukaemia

Age of children I: Mean 4.9 yrs (SD 2.5);

C: 7.8 yrs (SD 5.1)

Ethnicity not stated

Intervention: A two-stage process was used for one trial. 1. Written parental consent was sought for the induction phase of the trial during which all patients received the same induction chemotherapy. Written consent (4 weeks later) was then obtained for randomisation to one of four therapeutic regimens.

Comparator: Parents of children in the other three trials did not receive the staged approach. No further details provided.

A high possibility of selection bias due to lack of randomisation and no reported process for selecting individual participants. The intervention was not implemented in a standardised way. Due to poor reporting the risk of contamination was unclear. There was a particular risk of performance bias.

Coyne et al. (2003)[22] United States

Cluster randomised controlled trial

Adult cancer patients

Three trials (C9741; E1594; E2197)

I: n = 78;

C: n = 129

Breast (85%) and lung cancer patients

I: 92.3% female;

C 90.7% female

I: Mean 53 yrs;

C: mean 53 yrs

I: 94% white;

C: 92% white

Intervention: Easy to read version of the original written consent document (different for each of the three trials). Changes included text style, page layout, font size and vocabulary. Content was not altered. Readability was seventh to eighth grade level and length was 16 pages.

Comparator: Original consent document (different for each of the three trials). E1594: 4 pages long and fourteenth grade reading level. C9741 and E2197: 7–8 pages long and twelfth to thirteenth grade reading level.

This was a randomised study. The unit of randomisation was at the institutional level and this was maintained for the statistical analysis. However, due to poor reporting it is unclear whether the study was properly designed to protect against selection bias. The design appeared to protect against contamination as only one consent statement was used at an individual centre.

Donovan et al. (2003)[19] United Kingdom

Randomised controlled trial

Adult cancer patients

Single trial

I: n = 75;

C: n = 75

Prostate cancer patients

100% male

Age not stated

Ethnicity not stated

Intervention: Nurse conducted information appointment with the patient to recruit to the trial.

Comparator: Urologist conducted information appointment with the patient to recruit to the trial.

A good quality RCT: appropriate randomisation, concealed allocation, at least 80% of patients considered at follow-up and ITT analysis conducted. It is possible that contamination between the two groups and performance bias may have influenced the findings.

Donovan et al. (2002)[20] United Kingdom

Before and after

Healthcare professionals

Single trial

Baseline: n = 30;

I1: n = 45;

I2 n = 67;

I3: n = 83;

I4: n = 155

Prostate cancer patients

100% male Age not stated

Ethnicity not stated

Intervention: Three successive documents§ regarding how best to recruit patients to the trial were circulated to recruiters followed by a training programme. Consent to randomisation was measured at baseline and following circulation of each document.

This is an uncontrolled study therefore there is a risk of factors other than the intervention influencing patient participation.

Fleissig et al. (2001)[21] United Kingdom

Nonrandomised controlled study

Healthcare professionals and adult cancer patients

Forty trials

I: n = 135;

C: n = 130

10 different cancers

I: 72% female;

C: 72% female

Age range 19–65 yrs

I: 58% 45–64 yrs;

C: 50% 45–64 yrs

Ethnicity not stated

Intervention: Patients completed the Patient Preferences for Information Questionnaire, Patient Attitudes to Trials Questionnaire and Spielberger State Trait Anxiety Inventory prior to consultation with their doctor. Doctors were then provided with each patient's completed questionnaires (only the first 2 questionnaires) prior to their consultation during which consent was sought for a specific trial.

Comparator: Patients completed the same questionnaires prior to consultation with their doctor. Doctors were not provided with this information prior to their consultation with individual patients during which consent was sought for a specific trial.

A high possibility of selection bias: only the order in which doctors conducted intervention and control group consultations were randomised (in blocks of 5 patients). The process by which patients were selected for inclusion was not reported. There was a high possibility of contamination as the same doctors were involved administering the experimental intervention and the comparison. The intervention was not implemented in a standardised way. The process of completing the questionnaires may have influenced patient decision-making in both groups.

Gross et al. (2004)[25] United States

Controlled observational study

System level

Global target (National Cancer Institute phase II and III Clinical Trials Cooperative Group trials)

I: n = 4569;

C: n = 20,443 (2,440 were in phase II trials)

Breast, colon, lung and prostate cancer patients

Sex not stated

Age not stated 89% white

Intervention: Four states (Illinois, Louisiana, Virginia, New Jersey) that enacted legislation or developed a co-operative agreement with health insurers in 1999 to cover clinical trial patient care costs (coverage states).

Comparator: 35 states that had not enacted any policies to cover clinical trial patient care costs by the end of 2001 (non-coverage states)

The baseline enrolment rate was statistically significantly higher in intervention group than the comparator group introducing the risk of regression to the mean. Lack of enforcement in the intervention group and behaviour of physicians in the comparator states to compensate for lack of coverage could have had an influence.

Paskett et al. (2002)[23] United States

Nonrandomised controlled study

Adult cancer patients, healthcare professionals

All trials available to patients in a given geographical area.

Total number of participants not stated

Breast and colorectal cancer patients

Majority female

Age not stated for I and C (mean age, which was reported by time period of recruitment and cancer type ranged from 62 to 75 yrs)

75% white

Intervention: There were four elements: 1) a rapid tumour reporting system, 2) a nurse facilitator responsible for alerting physicians about appropriate clinical trials for their patients, 3) a quarterly newsletter about cancer treatment and clinical trials targeted at physicians and 4) a health educator who provided community-based education about screening and treatment and trained lay health educators. Implemented in five rural counties in North Carolina.

Comparator: No intervention in five rural counties in South Carolina.

The risk of selection bias is unclear. Data on patient trial participation were obtained from medical records; however it was unclear how specific cancer patients within regions were selected or whether all cases were detected. The study was susceptible to contamination: improving participation of patients in all rural areas was a major focus of the Community Clinical Oncology Program (CCOP) and both geographical areas had active CCOP physicians.

Simes et al. (1986)[26] Australia

Randomised controlled trial

Adult cancer patients and healthcare professionals

Thirteen trials at a single oncology unit

I: n = 28;

C: n = 29

8 different cancers

I: 82% female;

C: 62% female

I: mean 56 yrs (31–63 yrs);

C mean 55 yrs (40–74 yrs)

I: 96% white;

C: 100% white

Intervention: Uniform policy of total disclosure of all information relevant to the trial to the patient. There was an opportunity to ask further questions. Information was provided verbally and in a written consent form.

Comparator: Information about the aims, anticipated results and potential toxicities of treatment were provided with details of treatment provided at the discretion of the consultant. There was an opportunity for the patient to ask questions. Verbal consent was obtained.

This was a randomised study though it was not possible to assess from the information reported whether the method of assignment was truly random and whether it was concealed. There was a high possibility of contamination as the same doctors were involved in delivering the experimental. Attempts were made to establish whether the intervention and comparison were standardised across patients though it was not possible to establish whether the method used was sufficiently rigorous.