‘No longer has eczema’ was the primary reason for exclusion at the screening stage prior to the mail-out (
n = 66) and for parents declining participation (152 of 246 [61.8 %]). A large proportion of those identified for the mail-out were also excluded because they had not been diagnosed by a healthcare professional (
n = 39). Thus, in similar populations where diagnosis may be uncertain or transient, such as patients with childhood asthma, it should be expected that a large proportion of potentially eligible patients will not have a confirmed diagnosis or will no longer have the condition, as the Read codes in electronic medical records cannot be relied upon to exclude such patients. Accounting for this should aid calculations of return and recruitment rates for trials that employ this method of recruitment. The specificity of searches for potentially eligible participants can probably be improved by adding search criteria such as evidence of active disease through a recent relevant prescription, as in the Bath Additives in the Treatment of Childhood Eczema (BATHE) study [
31]. The decision regarding our original, narrower age range for eligibility was made on the basis that we thought older children were more likely to have an established treatment regimen and therefore would be less willing to be randomised to a different emollient. However, a recruitment yield of 12.6 % for children between the ages of 3 and 5 years (after the age criteria were broadened) suggests that there are still many families with older children who have not yet found a preferred emollient. Extrapolating from this result, we speculate that it is possible that, had the upper age limit been 5 years from the start of the study, an additional 26 % (taking an average of an additional six children per month) might have been recruited over the full 10-month recruitment period. Despite evidence derived from survey literature that reminders are helpful in increasing response rates [
32], reminder invitation letters sent to non-responders in our trial yielded four participants. It may be that an interim of three months between the initial and reminder letters was too long. Telephone reminders may have been more generative, as indicated by the results of other primary care trials [
33,
34]. The normal care pathway of the child and regulatory restrictions are also an important considerations. Many participants were seen and recruited by nurses, but governance restrictions on who could complete the trial paperwork hampered recruitment by this route. When these restrictions were relaxed, there was a rapid increase in the number of patients being referred into the trial by health professionals other than doctors (Fig.
3). When designing trials, researchers should give consideration to who normally diagnoses and treats the condition in question. If this includes health professionals other than doctors, they should be included in the eligibility criteria if at all possible.