Participant Retention in Depression Studies

CenExel is proud to highlight the poster presented at the International Society for CNS Clinical Trials and Methodology (ISCTM) Annual Scientific Meeting, February 2022, Washington, DC.

ABSTRACT

Introduction: It is well established that major depressive disorder (MDD) randomized clinical trials (RCTs) struggle to retain participants, with a 37% mean dropout rate (Khan et al., 2000) and a 48% mean dropout rate among smartphone apps trials of the same indication (Torous et al., 2020). Poor retention can lead to biased study results, reduced power, lower internal validity, less generalizability, and higher study costs (Liu et al., 2018). Studies exploring MDD trial attrition have notable limitations, such as meta-analyses have been conducted solely on published studies (Rutherford et al., 2012) and a lack of exploring a wider spectrum of variables that potentially might correlate with retention (e.g., participant compensation and number of scales at each visit) (Rutherford et al., 2013). The goals of the current investigation were to obtain a more comprehensive understanding of reasons MDD participants withdrawal and subsequently recommend strategies to improve retention. This was accomplished by analyzing several Independent Variables theorized to be linked with participant attrition and some of which have not been previously empirically explored (e.g., number of study scales and participant compensation).

Methods: The Dependent Variable in the current investigation was the number of participants who prematurely discontinued their involvement in the MDD clinical trial post baseline visit per their decision or situation (e.g., withdrew consent or lost to follow-up), as opposed to leaving the study because of study requirements or matters out of their control (e.g., investigator discretion due to labs or adverse events). The number of participants who withdrew from the study were obtained from Closeout Report Forms submitted to Institutional Review Boards and sponsors at the conclusion of a trial. This study collected the forms from 55 fully completed outpatient MDD clinical trials across 5 US sites located in the west and east coasts. Of the 55 trials, 39 were placebo-controlled (PC; n=417 randomized participants) and 16 were open-label (OL; n=285 randomized participants). Independent Variables were factors hypothesized to have a relationship with subject-driven attrition and were obtained from each study’s protocol. These variables were the number of study visits, frequency of those visits (derived by dividing study visits by weeks), study duration in weeks, number of scales per visit, likelihood of receiving placebo, and participant compensation.

Results: Pearson correlation analyses revealed that, for MDD PC studies (regardless of site geographical location), retention was significantly correlated with less study visits (r=-.40; p=.01), less study scales (r=-.39; p=.02), higher participant compensation (r=.58; p<.001), shorter study duration (r=-.37; p=.02), and less likelihood of being randomized to receive placebo (r=-.38; p<.02). For MDD OL trials, retention was significantly associated with higher visit frequencies (r=.57; p=.02) and attrition with longer study duration (r=.53; p<.05).

Conclusions: The results of the current investigation can be applied in developing protocols and implementing study strategies with an eye toward managing attrition. For example, sponsors may consider tempering the use of numerous exploratory psychometric scales to potentially lessen the rate of participant dropout post-randomization. Sponsors may also choose to limit the duration of a trial or increase the frequency of visits as a means to improve participant completion. While research sites are responsible for preventing patients from enrolling into clinical trials whose sole motivation is monetary, the current investigation indicates that higher compensation for time and effort spent in a trial significantly reduces premature trial discontinuation, and thus, requires fair deliberation when developing an Informed Consent Form. Other applications from the current investigation’s findings to protocol and study management aimed at lessening attrition as well as study limitations will be discussed in the poster.

BACKGROUND

• Expeditiously completing major depressive disorder (MDD) clinical trials for effective treatment is crucial as this is the most common mental health ailment in the United States, affecting 17.3 million adult Americans per year (NIMH, 2017), and is the leading cause of medical disease burden worldwide (World Health Organization, 2017)
• Impeding antidepressant research is the substantial attrition of MDD trial participation, ranging from 21% to 54% (Kahn et al., 2000; Rutherford et al., 2010; Torous et al., 2020), including randomized controlled MDD adolescent (Rohden et al., 2017) and MDD smartphone intervention app trials (Deady et al., 2020)
• Participant withdrawal is problematic given that this leads to missing data, costly delays in study completion (Costello, 2016; Ross et al., 1999), potential premature entire trial termination (Gul & Ali, 2010), negative research site moral (Sullivan, 2004), and questionable validity of the findings (Gul & Ali, 2010; Kadam et al., 2016; Levine et al., 2015; Torous et al., 2020). Moreover, statistical measures aimed to fix poor subject accrual has been noted as bias (Leon et al., 2006; Rabinowitz & Davidov, 2008).
• Researchers have thus investigated reasons for MDD participant dropout, including high study visit, higher visit frequency, and placebo-controlled studies vs comparator trials (Rutherford et al., 2012, 2013)
• Studies exploring MDD trial attrition are fraught with limitations:
  • Meta-analyses are typically conducted solely on published studies (Rabinowitz et al., 2009)
  • Lack of exploring a variety of hypothesized dropout variables due to inadequate access to such data (e.g., participant compensation and number of measures administered in the study) (Rutherford et al., 2013)
• Cohen et al. (2021a) specifically surveyed MDD patients about their reasons for staying in clinical trials: being treated with respect from site staff, provided free study visit transportation, and having the Informed Consent Form carefully explained by study staff. However, this investigation relied on participant recall regarding their discontinuation.
• The goals of the current investigation were to obtain (a) a more comprehensive understanding of MDD study participant dropout reasons and (b) subsequently recommend retention strategies by analyzing MDD clinical trials of various designs (regardless of their publication status or efficacy findings) possessing a multitude of Independent Variables (IVs) theorized to be associated with attrition and some which have not been empirically explored previously (e.g., number of study scales and participant compensation).

METHODS

• Table 1 lists the research site locations who participated in the investigation and the number and type of outpatient CNS clinical trials analyzed from those sites
• The trials analyzed for the present study were conducted any time starting 2018 and must have been completed (Last Patient Out) by May 2020
• Table 2 lists the IVs and Dependent Variables (DVs) analyzed in this study
• Dropouts were operationalized as participants withdrawing post-baseline (a) as obtained from the Closeout Report Forms submitted always to Institutional Review Boards and sponsors at the completion of each clinical trial and (b) directly due to participant decisions or circumstances (e.g., withdrawing consent and lost to follow-up) rather than matters out of their control (e.g., investigator discretion due to labs and adverse events)
• The IVs were obtained from the study protocol and checked for accuracy by two separate persons at each participating site
• All data were entered into an Excel spreadsheet and transferred / analyzed via a statistical program (SPSS v25.)

RESULTS

• 55 MDD indication clinical trials (see Table 1) conducted at 5 research sites across the US were analyzed
• Bivariate associations between the IV and DV were assessed via Pearson correlation analysis for continuous variables and point-biserial correlation analysis for a mixture of binary and continuous variables
  • Note we use IV and DP terminology ‘loosely’ as the statistical analyses in this study were not regressions, but rather correlational
• Percent Dropout (and Total n Randomized / n Completed) across all participating research sites for MDD PC trials were 23% (417/320), for OL trials 25% (285/213), and for Extension trials 35% (49/32)
• Results indicated a significant correlation for PC MDD studies’ retention post-baseline and less study visits (r=-.40; p=.01), less study scales (r=-.39; p=.02), higher participant compensation (r=.58; p<.001), shorter study duration (r=-.37; p=.02), and less likelihood of being randomized to receive placebo (r=-.38; p<.02); see Figures 1, 2, 3, 4, and 5 respectively. IVs not reported here were non-significantly associated with the dropout DVs.
• For MDD OL trials, retention was significantly associated with higher visit frequencies (r=.57; p=.02) and attrition with longer study duration (r=.53; p<.05)
• The PC and OL findings are consistent in correlational direction across the study’s US sites, although definitively concluding these finds are consistent regardless of site location is difficult because the East coast sites had low sample sizes (of trials) compared to the West coast sites, thus impacting statistical power

DISCUSSION

• The below describes the current investigation’s findings applicable to developing protocols and managing study procedures aimed at enhancing participant retention:
  • Our data connect higher retention with shorter study duration and less visits. The former finding mirrors the data from OL MDD trials too. Taken together, these results confirm those found by Rabinowitz et al. (2009) and Rutherford et al. (2013). Protocol developers are thus encouraged to be purposeful when designing a study’s length and visit structure so that they balance the need for data collection and retention. If reducing study duration and visits is not possible, and given the finding from the current study as well as our previous survey of patients (Cohen et al., 2021a, 2021b) who reported the importance of compensation to retention, researchers may also consider compensation ‘bonuses’ for completion of partial milestones to help reduce attrition.
  • Trials administering more psychometric scales in the current analysis tended to have higher dropouts. This is consistent with observatory data reported by our site Investigators, as participants can become frustrated completing multiple study measures, often perceived by the participants as asking redundant questions. Protocol developers should consider what instruments are necessary to collect the trial’s primary and/or co-primary endpoint data, as opposed to employing multiple exploratory and secondary measures that may be associated with higher study attrition.
  • Participants may withdraw when there is a greater likelihood of receiving a placebo. Similarly, Rutherford et al. (2012) found that participants with MDD show higher retention rates in OL (comparator) vs PC trials, presumably because participants randomized to placebo arms more likely experience continued depressive symptoms and thus seek further treatment outside of the study. While research site staff must cautiously balance therapeutic rapport so as not to heighten placebo response and potentially lessen participant dropout (Wampold, 2018), this can be managed by ensuring participants that their MDD symptoms reported during the study will be managed through the site’s after-trial referral and/or treatment services.
• Current study LIMITATIONS: (a) The low number of US sites (n=5) whose studies were analyzed may reduce the generalizability of the current findings; (b) new findings reported in this poster (e.g., visit frequency and number of scales) should be further investigated for confirmation; and (c) a reminder that correlational data is not causational and other factors may have contributed to the current results.

Elan A. Cohen,¹ Howard A. Hassman,¹ David P. Walling,² Vera M. Grindell,² John G. Sonnenberg,^3,4 Katarzyna Wyka,⁵ Brett A. English,^1,2 Jaclyn M. Lobb,¹ Djouher Hough,¹ Cassie L. Blanchard,¹ and Larry Ereshefsky^1,2,6

APEX Innovative Sciences – Hassman Research Institute¹; APEX Innovative Sciences – Collaborative Neuroscience Network²; Uptown Research Institute³; Northwestern University Feinberg School of Medicine⁴; The City University of New York, Graduate School of Public Health and Health Policy⁵; Retired Professor, The University of Texas⁶