Reasons for unsuccessful recruitment of children with atopic dermatitis in primary care in the Netherlands to a cohort study with an embedded pragmatic, randomised controlled open-label trial: a survey


A total of 367 patients were included in the cohort of the Rotterdam Eczema Study, of whom 209 patients were recruited throughout GP practices and were potentially eligible for the trial. Only 32 patients were ultimately selected for the trial and randomly assigned to a treatment arm. It was not possible to enrol enough participants despite reminders and additional information about the study procedure throughout the follow-up. Most of the participants did know what to do when they had a flare-up (92.8%). The majority of the patients in the cohort who completed the survey experienced a flare-up (56.5%); the main reason for failure to enrol a sufficient number of patients in the trial was that cohort participants did not contact their GPs when they had an AD flare-up. When participants did contact their GPs, most of them did not meet all the inclusion criteria for participation in the trial. The majority of the patients treated the flare-up themselves.

Strengths and limitations

A strength of the survey was that it gave us an understanding of the patients’ behaviour when they experienced a flare-up. These findings have let us make suggestions that may lead to better and more effective ways of recruiting trial participants in future studies in primary care. The low rates of trial participation may be attributed to selection bias, as patients with an understanding of their condition and who were more proactive in starting treatment were more likely to be enrolled in the cohort and therefore did not join the trial. A limitation of the survey is that we did not ask in depth why participants did not contact their GP. This makes it more difficult to comprehend the main reason for the trial’s failure fully.

Comparison with existing literature

Recruiting patients for RCTs in primary care is known to be difficult in a variety of ways.7 8 One of the interventions suggested by the Cochrane evaluation of methods to increase recruitment to randomised trials is using an open-trial design; this was also part of our study design.8 Furthermore, the Cochrane review suggests telephone follow-ups for those who do not reply to a postal invitation but our participants got a reminder in the weekly online survey to call their GP if their AD got worse. In retrospect, telephone reminders would probably have been more effective for our study. A recent study by Knapp et al found that multimedia information only (eg, animations and videos) increased the trial recruitment rate in children and young people compared with participant information sheets for trial recruitment.12 We used printed participant information only. However, the children in our study were already participating in a cohort study but maybe would have been more able and more willing to participate in the trial if multimedia information had been given.

We used a case-finding method in our study to recruit patients through general practices. When it turned out that inclusion for the cohort was behind schedule, we started recruitment through social media and a newspaper.6 Research by van der Worp et al showed largely comparable samples for recruitment throughout the media versus case-finding.13 That study was carried out in a comparable setting in Dutch general practice. Baker et al found that paid and unpaid social media recruitment could be an efficient tool that can potentially assist recruitment to clinical trials in AD.14 One solution could be to recruit solely through social media and at the same time increase the sample size of the cohort so that it would finally include more patients in the trial. However, to get a comparable patient selection, inclusion and exclusion criteria should be properly decided. To ensure that selection bias is minimised, it should be verified that the patient characteristics in primary care and open recruitment are identical. It should also be possible to overcome the logistical issue with medical supervision. When patients are recruited via their own GPs, the GP must have confirmed their willingness to take part in the research, and is responsible for the patient’s treatment and management. The GP can provide prescriptions and is the point of contact if the treatment is not effective. When patients are recruited through social media, that responsibility needs to be transferred to a physician in the research team.

A good example is the Panoramic trial of Butler et al. This was a nationwide, multicentre, primary care, open-label, multigroup, prospective, platform adaptive trial of early treatments for COVID-19 in the UK. They successfully included more than 10 000 patients.15 They included patients not only via the central trial team but also via hubs; these included GP Sites, Community Trusts and other health service providers, including government agencies for example, the UK Health Security Agency. A medically qualified professional, research nurse, nurse prescriber or prescribing pharmacist from the hub was able to complete all recruitment procedures, screening, baseline, informed consent and eligibility reviews. Furthermore, they also could provide the patient with the medication being studied. Although this trial had a larger budget and studied a potentially deadly disease with a high impact, it could be an interesting option to use the structure of hubs with medically qualified professionals to recruit patients for trials nationwide.

In our study, participants were told what to do in the event of an exacerbation of the AD in order to be included in the trial (92.8%), but a substantial number of patients did not contact their GPs when they experienced a flare-up (69.2%). Most of them started treating the flare-up themselves. This meant that participants did have a TCS at home, whereas the protocol stated that participants had to hand in any TCS they had at home during the cohort baseline visit if they did not use a TCS at baseline. They did not have to hand in the medication if they were doing maintenance therapy or treating a flare-up. Based on our baseline data, 49% of the children used a TCS at baseline, so this substantial group of participants probably had a TCS at home.16 One of the reasons for not contacting their GPs could be that if patients experienced a flare-up, it was probably more convenient to start a TCS they had at home than schedule a consultation with the GP. Or maybe they reconsidered participation when a flare-up of the AD occurred. In addition, an increase in AD symptoms might not have been noticed as a flare-up by parents/patients because it was assumed to be typical fluctuation of a well-known disease.

Furthermore, our window of inclusion covered the beginning of the COVID-19 pandemic. It is known that the number of consultations declined substantially compared with pre-pandemic levels.17 It was stated in the survey that COVID-19 was one reason why patients could not visit the GP when experiencing a flare-up. It is likely that patients did not want to burden the healthcare system unnecessarily during the pandemic or that the appointment was scheduled after too many days. This could have led to higher rates of patients who started treating the flare-up themselves.

Also, the high burden for trial participants could be a reason for failure to enrol.18 The burden of participating in our trial consisted of a consultation with the GP and three home visits.6 Patients had to contact the practice themselves to arrange an appointment with their GP. They might have had to wait 1 or 2 days before they could arrange an appointment, especially during the COVID-19 period. Additionally, the multicentre design resulted in one or only a few patients from each participating GP practice. This could have led to less awareness among the GPs and/or GP assistants about what to do when a participant was experiencing a flare-up. This was also mentioned in the survey. Moreover, the consultation with the GP was needed for going through the inclusion and exclusion criteria of the trial and objectively assessing the severity of the AD; this could also be one of the barriers that hampered contact with the GP during a flare-up. This barrier could be resolved by transferring those responsibilities to the research team and making them also available during out-of-office hours. In addition, digital photographs could be used to remotely assess the severity of AD. Studies with these methods are promising.19 20

Because their AD was mild or severe rather than moderate, the majority of patients who contacted their GP did not satisfy the trial’s inclusion criteria. One reason for this could be that our eligibility criteria were too narrowly defined. Narrow eligibility criteria and an overestimation of prevalence are known reasons for unsuccessful recruitment.18 21 We selected children with moderate eczema because we did not want to overtreat or undertreat them with the intervention and control treatments. Another reason for poor recruitment could be that we overestimated the prevalence of moderate AD in children. However, the baseline data of the cohort showed that mean AD severity was moderate on the POEM scale, so this does not seem to have been the problem. The survey showed that 56.6% of the participants in the cohort experienced a flare-up. If we had been able to recruit the cohort sample size through the GP practices, it would have been feasible to reach the desired trial sample size given the numbers and severity of AD cases in the cohort.

Implications for future research

We would like to present some suggestions for an improved design and a more successful way to answer the important research question of whether starting with a potent TCS during a flare-up of AD is more efficacious than starting with a low-potency TCS in children in primary care. See the summary in box 1.

First, responsibility for the patients’ AD should be transferred from their own GP through hubs to a physician in the research team or to a medically qualified professional such as a research nurse, as shown by Butler et al.15 Second, we consider the recruitment method. As discussed above, our own study and the literature show social media recruitment to be an effective way of including patients.13 14 Third, we suggest assessing the severity of AD digitally; the literature has shown that this is a promising alternative.19 20 Finally, we recommend implementing a direct medication delivery service to patients to eliminate the need for pharmacy visits. However, the GP or study team should verify the inclusion and exclusion criteria before randomisation. Eliminating the need for the patient to schedule an appointment with their doctor and visit the pharmacy makes it possible to reduce the time between the onset of the flare-up and the start of the randomised treatment. However, these recommendations are not absolute and depend on several factors, including the available budget, the condition being studied and the target patient group. We advise conducting a feasibility study to test a design that incorporates these modifications.

Box 1

Summary of recommendations for future research in children in general practice with AD

Recommendations for future research

Conduct a feasibility study

Choose inclusion criteria carefully

Responsible physician/nurse should be in the research team to supervise the treatment

Assess severity of AD digitally (via photos)

Deliver randomised medication to the home

Patient recruitment through social media or mobile research team

Telephone reminders for cohort patients

This post was originally published on