Hypothesis / aims of study
Only around half of all clinical trial succeed in recruiting enough participants to meet the originally specified sample size. In case of problematic recruitment, other strategies may be chosen, such as recruiting participants through social media or lay press. However, applying a different recruitment strategy may lead to sampling bias. This may impair the interpretation of the intervention outcomes and the generalizability of the study results.
We conducted a Randomized Controlled Trial (RCT) comparing an app-based treatment for women with urinary incontinence with usual primary care. We recruited both incident and prevalent cases through general practitioners, but experienced sincere problems with recruitment. We therefore decided to expand to (social) media approaches. To allow a proper interpretation of the generalizability of the outcomes, we analyzed the impact of recruitment strategy on the baseline characteristics of the participants in this RCT.
Study design, materials and methods
We are performing a mixed-methods study on the impact of a mobile application for urinary incontinence in adult women. Part of the study is a pragmatic randomized-controlled non-inferiority trail. Patients were only recruited through GPs from July 2015 to June 2018 (both incident and prevalent cases) and through both (social) media and GPs from November 2017 to June 2018.
The (social) media campaign consisted of interviews in regional newspapers spread through LinkedIn, Facebook and Twitter, interviews on national and regional radio and local broadcasting and directed advertisements on Facebook in the study region.
We collected characteristics through an online questionnaire, and during a baseline visit, consisting of medical history taking and a physical examination. The questionnaire assessed impact of incontinence symptoms (ICIQ-UI-SF questionnaire), quality of life (the EQ-5D-5L including the EQ-VAS) and disease specific quality of life (ICIQ-LUTS-QoL).
Analyses were performed on complete cases without imputation of missing data, because the number of missing values on baseline was expected to be limited.
Baseline characteristics of subjects in the three recruitment groups (incident cases, prevalent cases, social media recruitment) were compared using ANOVA or Kruskal-Wallis tests, depending on distribution of the data, for continuous data and with Chi-square tests for categorical data. A Bonferroni correction was applied to control for multiple comparisons. In case of a significant difference, a post-hoc test was performed to identify which recruitment groups differed.
Two-hundred-sixty-two subjects were randomized of which 256 (98%) subjects had complete baseline data and were included for this analysis. Of these, 120 were incident cases, 28 were prevalent cases and 108 were recruited through (social) media. Univariate between group comparisons showed only difference for age (Table 1, F=9.0, df =2, p<0.001). Post-hoc analyses showed that patients recruited through (social) media were significantly older than incident cases (mean difference 6.1 years (95% CI 2.5 to 9.7)) but that prevalent cases did not differ from the other two groups. No differences were found between groups in BMI, gynecological history and status, incontinence complaints, duration of incontinence and quality of life.
Interpretation of results
In this RCT on treatment of female urinary incontinence we found that samples recruited with different strategies were to a large extend comparable. Women included through the social media campaign were considerably older than women invited through general practices. However, this difference in age was not reflected in differences in age related outcomes such as symptom severity, type of urinary incontinence, menopausal status and quality of life.