Recruitment of low-income pregnant women into a dietary and dental care intervention: lessons from a feasibility trial

Difficulties in carrying out research in low-income and vulnerable urban communities are increasing due to issues of safety and setting-related factors (i.e. political instability) [1]. However, the challenges incurred with and suggestions for the solution of these difficulties are not often reported in scientific papers.

There is a wide recognition that recruitment of participants in research is a vital element for study success, particularly in clinical trials. However, an overwhelming number of clinical trials do not meet recruitment and enrolment targets [2]. A systematic review published in 2015 investigating causes of unsuccessful trial accrual found that 19% of trials had either an early termination due to recruitment failure or completion reaching only 85% of the expected enrolment target [3]. It has been estimated that approximately 86% of clinical trials fail to meet recruitment goals within their specified timeframe [3]. Moreover, the literature suggests that trial timelines can potentially be doubled beyond initially planned recruitment periods due to low participant enrolment rates [4].

Recruitment shortfalls lead to delays and increased costs with fieldwork, early trial termination and potentially failure to draw conclusions [2]. Inconclusive clinical trials have economic, scientific and political implications, as clinical and public health decisions will not be based on the best evidence possible [4‐6]. There are also ethical implications when patients are exposed to potential risks and the research does not provide advances in scientific knowledge and hence related recommendations due to lack of statistical power [7]. Additionally, recruitment failure can negatively affect the motivation and engagement of stakeholders who are closely involved with the study, participants and investigators [8].

Methodological papers describe a wide range of factors which can contribute to recruitment success, including trial design and protocol development, selection of sites, staff engagement, realistic estimation of recruitment targets and study timeframe, enhanced participant contact and communication, financial incentive or compensation and convenience and reduced participation burden [9, 10]. Regardless of scientific efforts made to identify and tackle barriers and challenges with recruitment, the problem still persists. Therefore, researchers can still benefit from real-world lessons learned from the implementation of feasibility trials. These experiences can be further shared with peers working in similar settings or scenarios. Feasibility trials are paramount to test recruitment strategies, acceptability of study design and viability of eligibility criteria, thus consequently informing the progression of large-scale definitive clinical trials [11].

With the overarching aim of adding to this body of evidence, we used recruitment data from the IMPROVE feasibility trial to (1) describe the main challenges of recruiting women to the study, (2) describe most commonly observed factors related to ineligibility and non-participation of women and (3) discuss potential successful strategies for promoting recruitment and enrolment for the future full-scale clinical trial.

The IMPROVE study was a feasibility trial on calcium and vitamin D milk fortification and periodontal therapy for improving maternal periodontal health and metabolic and inflammatory profiles. The trial was registered in the ClinicalTrials.gov database (NCT03148483) and approved by the Ethics Committee of the Maternity School of the Federal University of Rio de Janeiro-Brazil (certificate number 1.516.656). Ethical approval was granted for the quantitative and qualitative data collection and intervention delivery.

A mixed-methods sequential exploratory design was applied consisting of two phases: a qualitative followed by a quantitative phase [12]. The qualitative data were collected and analysed to assist with the study design and help explain the results found in the quantitative phase.

The study protocol has been fully described elsewhere [13]. Briefly, a 2 × 2 factorial feasibility randomised controlled trial employed four intervention groups (without cross-over). Concealed randomisation, using a mixture of permuted block sizes, stratified by smoking status was performed remotely via an online system developed by Sealed Envelope Ltd. Adult pregnant women with periodontitis were randomly allocated into the following groups: (1) fortified sachet with vitamin D and calcium and powdered milk plus periodontal therapy during pregnancy, (2) placebo sachet and powdered milk plus periodontal therapy during pregnancy, (3) fortified sachet with vitamin D and calcium and powdered milk plus periodontal therapy after delivery, and (4) placebo sachet and powdered milk plus periodontal therapy after delivery. Given the nature of the intervention (periodontal treatment), full blinding was not applied.

Baseline data were collected up to the second gestational trimester after checking for participant eligibility to the trial, which included a dental screening for periodontitis (T0), with follow-ups at third trimester (T1; during the course of the intervention) and 6–8 weeks postpartum (T2).

The absolute recruitment yields were lower than expected (n = 70 women vs n = 120 women), but the recruitment rate per centre was higher than expected (an actual recruitment of 2 women/week vs the predicted rate of 1.7 women/week). A wide range of factors influencing recruitment into a dietary and dental care trial for pregnant women in a low-income area were examined. The major reason for ineligibility was gestational age above 20 weeks at the time of first prenatal care visit, which was a patient-related factor beyond our control. The initial plan was to recruit women during the first trimester to allow for early periodontal treatment initiation. The literature suggested that starting periodontal therapy after 21 weeks of gestation might be too late to reduce the inflammation that is related to the adverse pregnancy outcomes [22]. After auditing of the primary care data, the study inclusion criteria were amended to accept women at up to 20 weeks of gestation. Even though health services were free, late onset of prenatal care is common among deprived populations [23]. However, there were some external factors that might have exacerbated the delay in prenatal booking.

Rio de Janeiro faced Zika, dengue and chikungunya epidemics during the trial recruitment. Zika was declared a national public health emergency in 2015 [24]. In February 2016, the World Health Organization (WHO) declared Zika a public health emergency of international concern [25]. Although the epidemic was considered controlled in 2017, the potential causal relationship between Zika infection during pregnancy and foetal microcephaly, which was highly debated in the media, might have interfered with the early attendance of prenatal care, as some pregnant women may have restrained themselves at home. The long strikes and local instability and violence might have also contributed to the late onset of first prenatal visit in this population.

The demographic statistics of the population of pregnant women assisted by the health centre showed that the women were relatively young. Therefore, it was anticipated that applying a too-rigid definition of chronic periodontitis (which tends to develop with age) would result in a low recruitment rate. Therefore, periodontitis was defined as the presence of one or more teeth with at least one periodontal site with ≥4 mm of clinical attachment loss with the presence of bleeding on probing. The presence of bleeding on probing ensured the existence of local inflammation.

The recruitment estimation was based on a prevalence of periodontitis of 47% in low-risk pregnant women in Brazil [26]. In our study, we found 42% of eligible women with periodontitis. However, the prevalence of periodontitis in our population might have been higher, as 64 women were excluded due to extensive dental decay and were not further examined. These women might have also presented with periodontitis, which was not accounted for.

The qualitative data showed that, although women were positive to the study intervention, they faced financial and social support challenges which might have interfered with study participation. Informed by the public and patient consultations, the study protocol included a delayed periodontal treatment (after childbirth) instead of non-treatment to ensure that all participants would benefit from the study. This was sought to improve interest in taking part in the study. All women received milk and sachets (with vitamin D/calcium for the fortified group and placebo for the non-fortified group) during routine prenatal visits. In addition, all participants received milk surplus according to the number of children living in the household. We also offered financial compensation for the participants to cover transportation and meal costs.

According to the study recruitment flow diagram, there were few protocol-related factors for non-participation. Offering childcare at the health centre (e.g. a nursery) while pregnant women were attending the study visit might have prevented some withdrawal of consent during the initial study phase. However, this would have required additional staffing and venues, which were not feasible in the present study.

Several setting-related factors hindered recruitment. The initial plan was to include at least two study sites. To be eligible, health centres must assist a deprived population, be located in a relatively safe area, have available consultation rooms for the study team, offer dental care and be willing to subsidise part of the cost of the delayed periodontal treatment. Only two sites were deemed eligible, and only one agreed to participate. The health centre operates on a catchment area basis; therefore, we were not allowed to recruit women served by other health care providers to the study site.

Faced with the high levels of crime and violence in Rio de Janeiro, the safety of fieldworkers was also a concern. Fieldworkers were victims of an armed robbery and experienced near-miss events of theft. For the definitive trial, provision has to be made in the research budget for fieldworker training on how to respond to tense situations as well as insurance and a contingency fund to cover medical expenses and replacement of stolen equipment, if needed.

The present findings from this feasibility study should be interpreted in light of the following limitations. Our findings are based on experiences from researchers working at a single site located in a metropolitan region of Rio de Janeiro State. Therefore, our conclusions might not be fully generalisable to other contexts. Nevertheless, given that many countries in Latin America (such as Brazil, Colombia and Mexico) share distinguishable characteristics regarding organised crime [27], it is reasonable to assume that at least some of the issues faced by our research team might be similar to issues experienced by others at different sites, both locally and internationally. Unfortunately, evidence of these relevant obstacles to the conduct of research at primary health care centres in developing countries is lacking in the medical literature. We, therefore, encourage other researchers to follow this path, disseminating pragmatic aspects of research conduct with the aims of both improving future study design and establishing this as a good practice among the scientific community. The priority of this study was to evaluate the feasibility of recruitment, eligibility rate and acceptability of the study design and intervention and to identify barriers to enrolment to inform the large-scale trial. Contrary to most clinical trials, we wanted to design a recruitment protocol which does not maximize internal validity in detriment of external validity. Although the aim was to balance both internal and external validity, there were recruitment factors beyond the control and scope of the study. For example, we lacked resources to provide transportation to the study site; instead, we offered reimbursement for travel. Women could bring their children to the visits, but the research team could not offer support for childcare. Women reported competing priorities (e.g. care for other children), family issues (e.g. partners in prison) and problems with accommodation which resulted in missed appointments. However, the study lacked resources to assist women dealing with these issues. Also, we had to exclude women with extensive caries who were potentially eligible for the trial, as the study could not provide dental treatment for these women prior to randomisation.

Although we have conducted public consultations and collected qualitative information prior to study commencement to ensure that the study design was in line with stakeholders’ views (for example, inclusion of a delayed control group, provision of additional milk to the family and small financial compensation for the participants’ time), a participatory approach where study participants are engaged as partners and involved in all stages of the research was not applied. This could have led to a higher recruitment rate and potentially retention.

Finally, our qualitative data were based on a hypothetical invitation to a trial. What women reported they would do or intended to do could have been different from how they would react when faced with a real decision. In-depth face-to-face interviews with individuals who declined study invitation or withdrew after intervention commencement would have provided useful information for the large-scale trial. However, budget and capacity limitations prevented a detailed assessment of the characteristics of those who declined study participation.

Inequalities in metropolitan areas are growing worldwide, posing a challenge to national and local policymakers. This situation calls for actions and more research. However, the difficulties in carrying out research in vulnerable urban communities are increasing. Although many studies are being undertaken in low-income urban areas, the methodological challenges faced by researchers and potential suggestions on how to deal with such problems are not often documented.

Our data illustrate the complexity of overcoming recruitment and enrolment challenges for clinical trials in resource-limited settings. The lessons learned in this study (Table 3) on factors influencing recruitment and enrolment of low-income pregnant women will inform the delivery of a large-scale definitive trial and may benefit other researchers designing interventions on similar settings.