Despite the paucity of evidence, this review has highlighted a few approaches that could be considered when designing and implementing a study.
It was found that sociocultural specific interventions have the potential to increase recruitment when recruiting a particular racial or ethnic group. Also, simple strategies such as personalised letters, inclusions of questionnaires, telephone reminders and monetary incentives were seen be effective. Trials using a patient preference design found no improvement in recruitment, nor did those implementing monitoring visits to trial sites, using nurses instead of doctors to recruit or those using an internet data capture system as opposed to a paper based one. Placebo use was detrimental to recruitment rate.
Strategies proven to improve recruitment
Sociocultural factors are often cited as being important considerations in trial design and one trial here sought to overcome such barriers to trial participation among African Americans . African-American men were randomised to a control group or one of three increasingly intensive intervention arms that used different combinations of mail, telephone and African-American church-based recruitment. The racial, ethnic and language backgrounds of the research team were all similar to those of the potential study participants. Each of the interventions designed to address the main areas previously identified as barriers to the recruitment of minority groups to clinical trials. However, although the most intensive intervention was the most successful, it only managed a 1% greater recruitment yield than the control arm, and raises question as to whether such intensive approaches are worth the resultant extra numbers recruited.
Alternatively, the success of the trained Hispanic women  suggest the use of culturally related recruitment methods may be effective at encouraging enrolment. However, although the authors do acknowledge that their study is weakened by the absence of a trained non-Hispanic group, this approach does appear worth considering for trialists.
Despite these cautions, there does appear to be an advantage in using cultural-specific strategies to improve recruitment to trials involving specific ethnic groups.
Trials with 'open' designs also appear to benefit recruitment. The practice of 'blinding' is used by many drug trials as it guards against a number of biases, such as ascertainment bias. However, their use is not pragmatic in the sense that in the 'real' world doctors and patients know what treatment they give and receive. There may be a role, certainly in confirmatory trials, to design these as open randomised trials in order to increase their generalisability, but the number of trials where such a design is possible may be limiting.
The simpler recruitment approaches such as telephone reminders, where almost three times the number were recruited compared to the no-reminder group, and inclusion of a questionnaire also showed an increase in numbers recruited (18% of those invited compared to 13.2% invited without the questionnaire). Therefore, trialists might consider ensuring such straightforward methods are adequately budgeted in for in their trial grant applications.
Martinson et al found that monetary incentives improved response rates amongst adolescents, had a positive effect on their willingness to be contacted about future intervention, thereby increasing the potential numbers who could be recruited. As this strategy has been identified before as an effective way of improving response rates to postal questionnaires , if it can also improve recruitment it is a worthwhile for trialists' consideration.
The apparent effectiveness of the above strategies, however, is based only on single studies of each method (except the 'open' design, with only two trials) and their inclusion in future trials is very much dependent on individual trial design. Therefore, if suitable, it may prove worthwhile for researchers to consider the inclusion of one of the above strategies. In turn, if included as a randomised element of the study, such trials would be adding to the evidence on the effectiveness of the strategy.
Strategies that may potentially improve recruitment
There were a number of strategies identified that showed potential, but would benefit from further research to decide one way or another.
Looking at the person undertaking the recruitment, the difference between a nurse and a doctor was not statistically significant. However, in addition to establishing that nurses are no less effective as surgeons in recruiting patients, Donovan and colleagues found that the nurses were the more cost effective option. As cost is always an important factor, the use of staff other than the clinician could be considered in some trials.
Interestingly, although it is quite common for research staff to visit the trial sites in order to encourage recruitment, in the case of trial co-ordinators doing so, the one trial identified here found no evidence of any benefit . Given the large time investment in making such centre visits, it would be worthwhile repeating this single study to either confirm or refute its findings.
It was seen that personalised letters did improve response rates and thereby the numbers that could be potentially recruited, but the effect was not statistically significant . The authors acknowledge the need for the study to be replicated with a larger sample, but the method does sound promising and has the added benefit of being a relatively simple strategy to implement.
With trial administration procedures playing a large part during trial, a pre-trial internet system studied by Litchfield et al did not show any difference in the number of patients recruited compared to the usual paper method. However there were gains seen in efficiency with its use. These were seen in the time from last patient completing the study to the database being released, data being entered more quickly and queries being resolved earlier.
The above strategies demonstrated the potential to have an effect on recruitment but not one that was large enough to warrant a recommendation for their use. However, as with those that were shown to be effective, the evidence of the effectiveness of these strategies is very limited, with additional and larger trials incorporating them into their design necessary to allow definite recommendations to be made.
Strategies that do not appear to improve recruitment
A number of strategies did appear to be far less successful, such as sending an advance postcard before mailing a full recruitment pack , where although a higher rate of questionnaire return and completion was seen, it did not increase significantly increase the percentage randomised.
Within patient preference trials, patients may be placed within one of three groups according to their preferences: 1) patients with no strong preference and consent to randomisation; 2) patients with a preference but still consent to randomisation; and 3) patients who refuse randomisation and choose their preferred treatment . This is in an attempt to prevent the bias that can occur when patients do not receive their preferred treatment option and could make clinical trials more attractive to those patients who are apprehensive of randomisation. Alternatively, a partially randomised design, as done by Cooper et al, where patients' preferences are identified before randomisation, with all consenting patients randomised . As Torgerson and Sibbald state, patient preference trials are not meant to replace randomised trials, but to complement them, although measuring a patient's preference maintains all the advantage of a RCT, but has the benefit of allowing for interactions between preference and outcome to be assessed . However, Cooper et al discovered no increase in the number of patients randomised .
Aaronson and colleagues found their telephone based nursing intervention did not increase the numbers recruited. In fact the intervention group showed a decline in recruitment rates. However, it was effective in increasing some aspects of patients' awareness and understanding of important issues regarding participation in clinical trials, which may benefit participation in the long term.
These apparently non-effective strategies, however, are based on the evidence of single trials. It would therefore be more reassuring to have additional trials to back up this scant evidence before making definitive recommendations to not implement any of these strategies.
Poor recruitment to clinical trials is a major threat to their validity. Low recruitment leads to loss of statistical power and reduces the generalisability of the study to the wider clinical population. Many trialists use unproven methods to try and improve recruitment to studies. Ironically many trialists rely on before and after methods to assess the effectiveness of interventions. For example, Donovan and colleagues claim to have increased recruitment rates to a prostate cancer trial by the incorporation of qualitative methods into the recruitment process . However, this intervention was not subjected to a controlled trial and therefore could have been confounded by regression to the mean, temporal changes or simple selection effects. Therefore, it is crucial that recruitment methods are underpinned by the best possible evidence and this means randomised trials.
One aspect that did not appear in any studies suitable for this review was education or training. In one cluster randomised trial in primary care published after our search strategy , a significant increase in recruitment rates to its trial of exercise and manipulation for low back pain, was seen in practices where primary care staff had been educated in the active management of back pain. This was however, an unforeseen effect of the training package and not a randomised evaluation of it as a strategy. Trialists, therefore, in light of the apparent effect might consider offering a professional education package to clinicians in order to boost recruitment.
Interestingly, a method often used to encourage the healthcare professionals to recruit patients (as opposed to encouraging the patients to participate) is payment, which despite being used extensively has limited evidence . In line with the limited number found here, the authors failed to find any controlled studies comparing recruitment rates with and without the financial incentives.
It is acknowledged here that searching for such methodological based papers can be difficult and that we may not have identified all potential papers. We may have been able to broaden the scope of this paper if we had reviewed all papers reporting recruitment strategies and not only those using random allocation, but it is felt that this would have furthered the situation of reported successes of strategies being anecdotal and not based on evidence. It is also unknown how many randomised trials of strategies have not been published due to negative results. Publication would allow trialists to make more informative decisions as to which strategies to use, and would add to the current limited evidence base.