The collection of biospecimens and biomakers is increasingly recognised as an important component of large scale epidemiological research. Such data enhance substantially the range and nature of the research questions that can be addressed and allow joint consideration of environmental and genetic factors. However, the establishment of large scale biobanks requires consideration of pragmatic and cost factors, alongside traditional scientific and methodological issues
We found that a single mailed invitation to existing cohort study members to donate a blood sample, have physical measurements taken and have these included in a long term biobank for unspecified research purposes yielded a substantial number of participants, with 33% of those approached taking part. The randomised data indicate that response rates were one-third lower at existing commercial pathology centres than at collection sites established specifically for the purpose of the study, where potential participants were given specified appointment times and phone reminders to attend their appointment. There was no significant difference in response rates according to whether or not participants were randomised to be invited to attend in a fasting or non-fasting state or according to whether or not they received a reminder letter. Those residing in regional areas were more likely to participate than those in urban areas. No systematic differences in data quality and biospecimen yield were seen between the dedicated clinics and pathology centres.
We are not aware of any other studies that have made randomised comparisons of biobank participation rates according to type of collection site, so the evidence to date in this area is limited. There are several plausible explanations for the higher attendance rates at dedicated clinics. Studies researching the effect of invitation type on attendance for breast and cervical cancer screening have established that assigned appointments increased uptake of screening
[7–9] and this was the method of choice for recruitment into the UK Biobank
. A phone call reminder to follow up a written invitation is an additional effective method of increasing participation
[10, 11]. Furthermore, several studies have found that establishing a separate clinic in general practice for cancer screening and vaccinations is an effective way of increasing response rates
[8, 12, 13], however whether this finding is generalisable to recruitment to biobanks is unclear. The dedicated clinics were able to accommodate weekend recruitment, which was not possible through the pathology centres.
There is little research that can explain why people may be willing to participate in a biobank in some situations, but not others. Previous studies indicate that multiple methods of contact with participants leads to higher response rates in studies in general
[10, 11]. Other studies have found that multiple reminders are an effective way of increasing response rates to postal questionnaires
[14, 15]. However, in the current trial, a single reminder letter sent to half the cohort members randomised to attend a pathology centre in Parramatta did not have a significant effect on response rates. The reason for this is not known; potential explanations include the fact that only one reminder was given, that the reminder was also sent by mail, that study power was limited or that no genuine difference was present, in this setting.
Research has previously been undertaken to investigate whether individuals would be willing to participate in biobanking, using hypothetical scenarios
[16–19], however, data comparing the factors related to actual participation, are limited. The results from the present trial indicated that response rates were higher amongst participants in the regional area than in the urban area. A large multicentre US study found that enrolment site was the most important influence on consenting to donate biospecimens, with response rates varying between 40% and 100% over the 24 enrolment sites used in the study
. However, although it was noted that the sites represented a diverse range of urban and regional areas, the researchers did not indicate which locations yielded higher response rates. In contrast, another biobank study found no significant difference based on residential area, however this study measured willingness to participate and not actual participation
There are a number of potential limitations of the current trial that should be borne in mind. We used address details from the baseline mailout for The 45 and Up Study cohort, updated where participants notified The 45 and Up Study Coordinating Centre of a change of address. However, it is likely that some participants had moved, without updating their address details, in the one to three years since joining The 45 and Up Study; others may not have received their invitation or may have been unavailable to take part in the Link-Up Study for some other reason (e.g. holidays, illness). Although the randomised design means that this is unlikely to bias the main comparisons, it does mean that the response rate should be considered to be the proportion of people taking part in the biobank as a proportion of those sent invitations, not as a proportion of those who actually received an invitation. The trial only compared two models of biospecimen and physical measurement data collection. It is therefore not possible to separate different elements of the two designs. For example, it is not possible to establish what the separate contributions of the specific appointment time, weekend appointments and the phone reminders were to response rates, since these were all part of the general dedicated clinic protocol.
The trial examines response rates to biospecimen collection and physical measurements among people already participating in The 45 and Up Study. Similar to most cohort studies, The 45 and Up Study is not designed to be representative of the general population
. Rather, it aims to provide a large cohort of individuals consenting to comprehensive and long term follow up, with sufficient numbers in a range of population and exposure groups to allow for valid internal comparisons. The response rate to the main cohort was ~18%. The overall response rate to the Link-Up Study of 33% could be interpreted broadly as 6% of a general population study taking part; comparable to the 9% seen in the UK biobank, but less than other more localised community-based cohort studies (e.g. EPIC Norfolk, Framingham).
The Link-Up Study aimed to test potential strategies for large scale collection of biospecimens and physical measurements from The 45 and Up Study participants. While there was some limited promotion of the trial in the annual newsletter distributed to 45 and Up Study participants (via post or e-mail) and on the website of The 45 and Up Study, there was no media promotion of the trial or local publicity, or other strategies which could potentially improve local participation. This is because the trial sought to explore methods that could be used for The 45 and Up Study as a whole and there would be potential difficulties doing this across the ~800,000 km2 of NSW, for a prolonged period of time. It shows the response rate to a single ‘cold’ invitation to take part, with reminders. There remains considerable potential for improving response rates.
Dedicated clinics are considered the gold standard for collection of biospecimens and biomarker data for cohort studies, but are relatively expensive and time intensive to establish. Funding must be sourced for infrastructure such as premises and equipment and qualified staff must be hired and trained; these costs will be greater when sampling a geographically dispersed population. In addition to cost, there are logistical implications associated with establishing a clinic de novo. Once established, they can generally only accept participants for a restricted time period.
Collection through existing pathology centres has the advantage of being easier to implement, and having a broad network of collection locations that can be utilised fairly rapidly. Disadvantages are that pathology centres are less flexible and study design may need to be adapted to this; for example, large numbers of fasting participants, designated appointment times and weekend collection may not be able to be accommodated. While response rates were 50% higher at dedicated clinics, compared to commercial pathology centres, the total weighted costs in this trial were 110% higher per invitee, and 40% higher per final participant, at the dedicated clinic versus the pathology centres.
The 45 and Up Study cohort is spread across NSW, with over 50% in rural and regional areas. It would not be practical to set up dedicated clinics to serve the entire cohort, especially in sparsely populated areas. Hence, in this case, a mixed method of collection is likely to be the most practical.
It is generally assumed that more specialised staff, experienced in the specific type of data collection entailed, will collect data that is of higher quality than less specialised or experienced staff. However, the assumption has not, to our knowledge, been tested empirically in the setting of a biobank. The dedicated clinics were set up for the sole purpose of collecting blood samples for research, in accordance with the trial protocol. Conversely, the pathology centres provide a general commercial service, collecting a range of biospecimens for various purposes (predominantly clinical) and according to different instructions. The pathology staff relied on trial participants bringing the Pathology Request Form with them, which provided specific instructions to the phlebotomist. We found no material differences in the physical measurements taken between the dedicated clinics, with more specialised staff, and those at the pathology centres, and biospecimen yields were close to or exceeded targets for the vast majority of participants. This suggests that specialisation beyond the basic training received by pathology centre staff for the purposes of this trial did not appear to be accompanied by a discernable improvement in data quality. If rolled out on a larger scale, specific training of staff at every commercial pathology centre may be impractical and the incidence of protocol deviations could potentially be greater, as there would be less awareness of the study amongst staff in a broad network of pathology centres. Hence, the findings observed here may represent a best-case scenario.