Recruitment, retention, and adherence in a clinical trial: The Pediatric Heart Network’s Marfan Trial experience

Abstract

Background/Aims:

The Pediatric Heart Network Marfan Trial was a randomized trial comparing atenolol versus losartan on aortic root dilation in 608 children and young adults with Marfan syndrome. Barriers to enrollment included a limited pool of eligible participants, restrictive entry criteria and a diverse age range that required pediatric and adult expertise. Retention was complicated by a 3-year commitment to a complex study and medication regimen. The Network partnered with the Marfan Foundation, bridging the community with the research. The aims of this study are to report protocol and medication adherence, associated predictive factors and to describe recruitment and retention strategies.

Methods:

Recruitment, retention and adherence to protocol activities related to the primary outcome were measured. Retention was measured by percentage of enrolled participants with 3-year outcome data. Protocol adherence was calculated by completion rates of study visits, ambulatory electrocardiography (Holter monitoring) and quarterly calls. Medication adherence was assessed by the number of tablets or the amount of liquid in bottles returned. Centers were ranked according to adherence (high, medium and low tertiles). Recruitment, retention and adherence questionnaires were completed by sites. Descriptive statistics summarized recruitment, retention and adherence as well as questionnaire results. Regression modelling assessed predictors of adherence.

Results:

Completion rates for visits, Holter monitors and quarterly calls were 99%, 94% and 96% respectively. Primary outcome data at 3 years was obtained for 88% of participants. The mean percentage of medication taken was estimated at 89%. Site and age were associated with all measures of adherence. Young adult and African American participants had lower levels of adherence. Higher adherence sites employed more strategies, had more staffing resources, less key staff turnover, more collaboration with referring providers, utilized the Foundation’s resources and used a greater number of strategies to recruit, retain and promote protocol and medication adherence.

Conclusions:

Overall adherence was excellent for this trial conducted within a NIH-funded clinical trial network. Strategies specifically targeted to young adults and African Americans may have been beneficial. Many strategies employed by higher adherence sites are ones that any site could easily use, such as greeting families at non-study hospital visits, asking for family feedback, providing calendars for tracking schedules and recommending apps for medication reminders. Key lessons learned related to differences in adherence by age and race, site differences, the value of collaborative learning and partnerships with patient advocacy groups. These lessons could shape recruitment, retention and adherence to improve the quality of future complex trials involving rare conditions.

Background/Aims

Marfan syndrome is a rare, genetic connective tissue disorder that affects multiple systems. Aortic dissection is the leading cause of death.¹ From 2007 to 2011, the Pediatric Heart Network Marfan Trial randomized 608 children and young adults with Marfan syndrome to compare atenolol versus losartan in slowing aortic-root enlargement.² The trial was sponsored by the National Heart, Lung, and Blood Institute’s Pediatric Heart Network in partnership with The Marfan Foundation, the largest patient advocacy group for individuals with Marfan syndrome and related conditions (www.marfan.org).

The study design for the main trial has been published previously.³ In-person study visits were conducted at baseline, 6, 12, 24 and 36-months post-randomization. Echocardiograms, which provided the primary outcome measure, were performed at each visit. Heart rate measurement, important for study medication titration, was measured on a 24-hour ambulatory electrocardiogram recording device (Holter monitor) before, during and at the end of titration as well as at the time of in-person visits. Quarterly calls to assess participant well-being and adverse events were conducted at 8 time points between visits: at 3, 9, 15, 18, 21, 27, 30 and 33-months post-randomization. Each of these study activities had a protocol-specified window.

Recruitment, retention and adherence to the protocol and medication regimen are critical variables influencing trial outcome. Although recruitment and retention in clinical trials are difficult, and strategies to improve success in these areas are essential, it’s not easy to quantify the effectiveness of such strategies.^4–5 Conducting a randomized clinical trial in participants with a rare condition presents additional challenges.^6–7 A small pool of patients to which rigorous inclusion and exclusion criteria are applied further limit the available pool. In the Marfan Trial, the wide age range (6 months up to 25 years) required expanding recruitment to include both pediatric and adult centers and utilizing different strategies and messaging to reach diverse populations of parents, adolescents and young adults.

Recruitment for this trial was conducted in the shadow of animal studies that suggested a dramatically positive treatment effect of losartan in mice, believed to potentially extend to humans.⁸ Due to the animal work and encouraging results in a small cohort of pediatric patients,⁹ losartan gained increasing visibility within the Marfan community as a potentially extraordinary drug. Willingness to be randomized became an additional recruitment challenge as some treating physicians began prescribing, and patients increasingly requested, losartan.

Some protocol activities required participants to engage in procedures that were difficult or undesirable, such as wearing the Holter monitor to school. The calls and in-person visits were demanding for both participants and sites. The protocol and manual of operations provided a general outline to recruitment, retention and adherence, but sites were also able to create site-specific tools and strategies with Institutional Review Board or ethics committee approval. Sites were encouraged to develop innovative strategies and regular conference calls provided a forum to share best practices.

After publication of the trial results, the study committee reflected on what contributed to success and what could be improved upon in future studies. Our aims are to report the completion rates for study procedures as measures of protocol adherence; to report proportions of medication returned as a measure of adherence; and to describe the specific strategies employed to meet target enrollment and avoid participant withdrawal while promoting adherence. The strategies used in the Marfan Trial may provide valuable insight for trials facing similar challenges, particularly in research involving rare conditions.

Methods

The Marfan Trial protocol was prospectively reviewed and approved at all 21 participating sites by a duly constituted Institutional Review Board or ethics committee and conformed to the International Council for Harmonization Good Clinical Practice guidance. Written informed consent was obtained from each participant. The Marfan Trial was registered at www.clinicaltrials.gov (NCT00429364).

Study design

This analysis evaluated recruitment, retention, and adherence to protocol activities related to the primary outcome (change in aortic root z-score, corrected for body surface area, over three years). Recruitment was measured by the number of participants eligible and consented. Retention was measured by percentage of participants with 3-year outcome data. Protocol adherence was calculated by overall completion rates of in-person study visits, Holter monitors and quarterly calls, as well as completion rates within study windows. Medication adherence was assessed by the number of tablets or the amount of liquid in medication bottles returned.

A Marfan Trial Recruitment, Retention and Adherence site questionnaire was distributed to the principal investigator and lead coordinator at each site. Site investigators and coordinators were encouraged to confer with the trial team when completing the questionnaire, which was designed to compile information regarding staffing resources and strategies employed to enhance recruitment, retention and adherence.

Statistical analysis

Descriptive statistics were used to summarize trial recruitment, retention, and adherence, as well as questionnaire results. A paired t-test was used to test for differences in percent of bottles returned before vs. after May 2011, when strategies were discussed on a conference call. Bivariate regression models were used to assess predictors of protocol and medication adherence. The following predictive variables were included in the analyses based on demographic data collected: age at randomization, gender, race/ethnicity, previous medication usage, neurodevelopmental or psychiatric disorders, family history of Marfan syndrome and site. Predictors significant in bivariate modelling (p < 0.05) were included in a multivariable regression model.

Site adherence scores were assigned based on the completion rates for study visits, Holter monitors and quarterly safety calls, as well as bottle return rates. For each category, sites were assigned a score of 1 for low completion, 2 for medium completion and 3 for high completion, based on tertiles of completion relative to other sites. Scores were added over the four categories (maximum score 12); sites with scores 4–5 were considered “lower adherence” sites and those with scores 11–12 were considered “higher adherence” sites. Responses to the site questionnaire were compared descriptively between “higher” and “lower” adherence sites.

All analyses were conducted using SAS v9.4 (SAS Institute Inc., Cary, NC), and statistical significance was tested at level 0.05.

Results

Recruitment

Across participating sites, 1367 medical records were reviewed. Screening visits were conducted for 701 participants and 608 went on to be randomized.² Enrollment by site ranged from 7 to 57 with a median of 30 and mean of 29 participants per site.

Completion of in-person study visits/echocardiograms, Holters, and quarterly calls

The overall completion rates for expected study visits/echocardiograms, Holter monitors and quarterly calls were 99%, 94% and 96%, respectively (Table 1). The withdrawal rate was low (11% in each group), and 88% of all participants attained the primary outcome (echocardiograms 36 months after randomization). Most of the visits/echocardiograms, Holters and calls were successfully completed within the specified windows (89%, 84% and 89%, respectively).

Table 1:

Completion of in-person visits/echocardiograms, Holters and quarterly calls

Visit Completion Percentage, Overall and In-Window
Visit Completion	N	Mean Overall Completion (%)	Mean In-Window Completion (%)
6 Month Visit	594	99.7	92.4
12 Month Visit	589	99.3	88.6
24 Month Visit	575	98.6	87.4
36 Month Visit	549	98.2	89.5
Total	594	98.7	88.7

Holter Completion Percentage, Overall and In-Window
Holter Completion	N	Mean Overall Completion (%)	Mean In-Window Completion (%)
End of Up-titration Visit	586	96.6	94.4
6 Month Visit	591	98.0	85.5
12 Month Visit	580	97.1	84.0
24 Month Visit	567	95.6	82.7
36 Month Visit	544	91.7	80.9
Total	593	93.9	84.0

Quarterly Call Completion Percentage, Overall and In-Window
Quarterly Call	N	Mean Overall Completion (%)	Mean In-Window Completion (%)
9 Month Call	593	97.1	86.5
15 Month Call	589	97.8	91.3
18 Month Call	586	97.3	88.9
21 Month Call	581	97.3	90.5
27 Month Call	573	97.7	92.7
30 Month Call	563	97.5	90.6
33 Month Call	559	94.6	90.2
Total	598	96.5	88.7

Completion rates did not vary by treatment arm (atenolol versus losartan), but younger age at randomization, White vs. African American race and site were predictors for greater protocol adherence (Tables 2–4). Additional predictors of better adherence included a negative family history of Marfan syndrome for visit completion, female gender for Holter monitor adherence and history of neurodevelopmental disorder for quarterly calls.

Table 2:

Factors Associated with Visit Completion Percentage

Category	N	Mean Adherence Overall (%)	Bivariate Modelling p-value	Multivariable Modelling p-value	Mean Adherence In-window (%)	Bivariate Modelling p-value	Multivariable Modelling p-value
Total	594	99			89
Treatment Arm			0.41			0.23
Atenolol	294	99			90
Losartan	300	98			88
Age at Randomization			0.23			0.01	0.008
6 months to < 8 years old	213	99			92
8 to < 13 years old	157	98			86
13 to 18 years old	136	99			89
>18 years old	88	98			84
Sex			0.58			0.59
Male	357	99			89
Female	237	99			88
Race*			0.01	0.007		0.36
White	513	99			89
African American	45	96			86
Asian	16	100			94
American Indian or Alaskan Native	3	100			100
Native Hawaiian or Pacific	3	100			100
Islander
More than one race	14	100			91
History of Psychiatric Disorder			1.00			0.84
Yes	39	99			89
No	555	99			89
History of Neurodevelopmental Disorder			0.32			0.54
Yes	116	99			90
No	478	99			88
Family History of Marfan			0.01	0.01		0.35
Unknown	23	100			91
Yes	352	98			88
No	219	100			90
Clinical Research Site†			0.003	0.005		<0.001	<0.001
	6 – 56	91.7–100.0%			62.5–100.0%

^*due to small sample sizes, the p-value reflects only the comparison of Whites to African Americans

^†range of sample sizes and mean adherences across the 21 sites

Table 4:

Factors Associated with Quarterly Call Completion Percentage Significant covariates only (p-value<0.05).

Category	N	Mean Adherence Overall (%)	Bivariate Modelling p-value	Multivariable Modelling p-value	Mean Adherence In-window (%)	Bivariate Modelling p-value	Multivariable Modelling p-value
Total	598	96			89
Age			0.16			0.02	0.06
6 months to < 8 years old	215	97			89
8 to < 13 years old	156	98			90
13 to 18 years old	137	96			90
>18 years old	90	94			83
Race*			0.002	0.008		0.03	0.03
White	516	97			89
African American	46	90			82
Asian	16	100			90
American Indian or Alaskan Native	3	100			100
Native Hawaiian or Pacific Islander	3	100			96
More than one race	14	100			97
History of Neurodevelopmental Disorder			0.07			<0.001	<0.001
Yes	116	98			94
No	482	96			87
Clinical Research Site†			<0.001	<0.001		<0.001	<0.001
	6 – 56	83 – 100			72 – 99

^*due to small sample sizes, the p-value reflects only the comparison of Whites to African Americans

^†Range of sample sizes and mean adherences across the 21 sites

Medication adherence: Bottles returned, percentage of medication consumed

The mean percentage of bottles returned was 67% and varied significantly by age (higher for younger; e.g., 74% age<8 years vs. 56% age>18 years; p=0.002), race (higher for Whites vs. African Americans: 68% vs. 53%; p=0.02) and site (range 20–96%; p<0.001) (Supplemental Table 1). In May 2011, interim adherence data was shared with study staff and higher adherence sites shared their strategies for bottle return. This simple intervention resulted in an increase in the mean percentage of bottles returned from 63% to 68% (p=0.015). At study end, the proportion of subjects who returned at least 60% of medication bottles was 65%.

We estimated the mean percentage of medication taken to be 89% based on residual amounts in returned bottles. The percentage of medication taken varied significantly by site (p<0.001), race (p<0.02) and age (p<0.002), but not by treatment arm, gender, history of neurodevelopmental disorder, history of psychiatric disorder or family history. By site, the adjusted mean (95% confidence intervals; CI) percentage of medication taken ranged from 78 (CI 72, 84) to 100 (CI 88, 112).

Clinical research site questionnaire

All enrolling sites were invited to complete a site questionnaire. Sixteen sites returned fully completed questionnaires, one returned a partially completed questionnaire and four did not return the questionnaire. Lower and higher adherence sites were differentiated based on the completion of study visits/echocardiograms, Holter monitors, quarterly calls and bottle return rates (Table 5). Of the 608 participants enrolled, 534 (88%) were recruited from the 17 sites that responded to the questionnaire (Table 6). Three non-responding sites were in the middle tertitle, while one was in the lower adherence tertile. Mean recruitment from the non-responders was 18.5, compared to 31 participants for responders. There was considerable variability in staffing resources within clinical research sites. Higher adherence sites tended to have greater staffing resources (3.5 study staff versus 2.5 at lower adherence sites). Interestingly, 4/6 higher adherence sites had no turnover of key staff for the entire study [principal investigators and study coordinators]. One lower adherence site had three different principal investigators and three different study coordinators.

Table 5:

Site Adherence Scores Based on Visits, Holters, Calls, and Bottles Returned

Site	# (%) Enrolled	Visits	Calls	Holter	Bottles Returned	Adherence Score
A	35 (6)	1	1	1	1	4
B*	7 (1)	1	1	1	1	4
C	37 (6)	1	1	1	1	4
D	19 (3)	2	1	1	1	5
E	30 (5)	1	2	1	1	5
F	9 (1)	1	2	1	1	5
G*	22 (4)	1	2	2	2	7
H*	9 (1)	2	2	1	2	7
I	39 (6)	2	1	2	2	7
J	37 (6)	2	2	2	1	7
K	57 (9)	3	1	2	2	8
L	13 (2)	3	1	2	2	8
M*	36 (6)	2	2	2	2	8
N	33 (5)	2	2	3	2	9
O	27 (4)	2	3	2	3	10
P	17 (3)	3	3	2	3	11
Q	26 (4)	2	3	3	3	11
R	20 (3)	3	2	3	3	11
S	49 (8)	3	3	3	3	12
T	54 (9)	3	3	3	3	12
U	32 (5)	3	3	3	3	12

^*Did not return site survey

Lower Adherence Sites

Higher Adherence Sites

Sites were assigned adherence scores based on the completion of study visits/echocardiograms, Holter monitors, and quarterly safety calls, as well as bottle return rates. For each category, sites were arranged by tertiles of completion rates and assigned a score of 1 for low completion, 2 for medium completion, and 3 for high completion. Scores were added over the four categories and total scores ranged from 4–12; those with scores 4–5 were considered “lower adherence” sites and those with scores 11–12 were termed “higher adherence” sites. Of note, one of the lower adherence sites (site B) did not respond to the questionnaire, and one (site F) responded to only a few questions. All higher adherence sites returned the questionnaire.

Table 6:

Marfan Trial Recruitment, Retention, and Adherence Site Questionnaire

Respondents	N	%
Marfan trial sites represented (core & auxiliary)	17/21	76
Trial participants recruited from responding sites	534/608	88

Recruitment Strategies Utilized By Sites
Strategy	N†	%
In-person recruitment from site’s clinic	16/16	100
Financial support for out-of-pocket research-related expenses	14/16	88
Direct mailings to potential participants	13/16	81
Collaborations with local physicians	12/16	75
Direct mailings to targeted healthcare providers	11/16	69
Speaking engagements, presentations, health fairs, etc.	11/16	69
Trial information posted on site’s website	7/16	44
Printed on-site recruitment materials	3/16	19
Site’s social media	0/16	0
Regional paid advertising (newspaper, radio, television, etc.)	0/16	0

Retention Strategies Utilized By Sites
Strategy	N†	%
Update participant’s contact information regularly	16/16	100
Utilize the participant’s preferred means of communication	16/16	100
Establish/reinforce expectations around study activities	16/16	100
Communicate with the participants regarding trial results	16/16	100
Expressions of gratitude from the study staff to the participant	16/16	100
Accessibility of the PI to the participant to promote trial engagement	16/16	100
Provide reminder calls, emails, texts day prior to study visit	15/16	94
Use of alternate contacts to prevent loss-to-follow-up	15/16	94
Advanced scheduling of study visits (typically prior to leaving clinic)	15/16	94
Initiate participant contact at clinical/non-study related visits	15/16	94
Financial support of participant’s out-of-pocket research visit expenses	15/16	94
After visit summary letters sent to primary physician after each visit	15/16	94
Enrollment letters sent to primary physician at enrollment	15/16	94
Customized study calendars to track participants’ trial activities	14/16	88
Periodic printed correspondence (letters, study newsletters, etc.)	12/16	75
Reminders/itineraries sent prior to study activities	10/16	63
Send greeting cards for birthdays, holidays, etc.	7/16	44
Certificate of study completion and/or service hours for participation	3/16	19

Study Medication Adherence Strategies Utilized By Sites
Strategy	N	%
Verbal education about when and how to take the study medication	17/17	100
Instructions to incorporate the study medication into the daily routine	17/17	100
Written education about when and how to take the study medication	16/17	94
Providing study-medication info cards (medication/emergency contact information)	15/17	88
Providing or advocating for the use of pill minder boxes, pill cutters/crushers	14/17	82
Additional resources, education, materials, and tools for low-adherence	14/17	82
Advocating use of reminder alarms	13/17	77
Use of apps to assist with medication tracking	7/17	41
Incentives for the return of study medication bottles	6/17	35
Other strategies self-reported by sites:
Frequent communication: calls, texts and emails	14/17	82
Pre-paid, pre-addressed mailers provided for bottle return	10/17	59
Personalized calendars	4/17	24
Reminder magnets for the fridge	1/17	6

^†N was adjusted from 17 to 16 since one site did not submit responses to this section of the questionnaire

Recruitment strategies

Clinical research sites did not differ in who handled recruitment, retention and adherence activities. Principal investigators, study coordinators, geneticists, and pediatric cardiologists were responsible for identifying potential participants and talking to them about the trial.

The Pediatric Heart Network’s Recruitment and Retention Plan and associated checklist (Supplemental Table 2) contains a wide variety of strategies to guide site recruitment and retention efforts for studies launched within the Network. This formed the basis for the Data Coordinating Center to develop recruitment templates, talking points, and letters to referring physicians; however, sites were able to choose, customize and develop their own site-specific materials as well. Ideas, tips and study materials were also shared during conference calls and biannual in-person meetings. Of note, participant compensation and incentives were determined at the discretion of the clinical research site. The site questionnaire (Table 6) did not collect site-specific compensation/incentive plans and only reported if incentives were offered.

All sites reported using internal rosters of patients from which to recruit with 81% mailing information directly to participants. Internal rosters came from clinics, referring providers and medical record queries. Sites sent mailers directly to healthcare providers (69%), collaborated with local providers caring for this population (75%) and used health fairs, speaking engagements and presentations to discuss trial information to providers and potential participants (69%). Social media (Facebook, Pinterest and Twitter) and paid advertising (newspaper, radio and television) were not utilized. Challenges identified by the sites were age (teens and young adults less likely to enroll), willingness to be randomized and distance from the clinical research site.

Recruitment strategies differed between higher and lower adherence sites. Higher adherence sites more often engaged in collaborations (mailing and calling) with healthcare providers (pediatric cardiologists, ophthalmologists, geneticists) and reported using a greater number of recruitment strategies such as speaking at patient advocacy events, conferences, and institutional grand rounds and leveraging The Marfan Foundation’s resources, educational materials and website.

Retention strategies

There were few differences in the types of strategies used for retention. As with recruitment strategies, higher adherence sites reported using a greater number of strategies. All sites reported a focus on building rapport and relationships with participants and engaging in continuous and positive communication with families, most often done by the principal investigator and reinforced by other study personnel. Strategies included managing expectations, communicating results, expressing gratitude for participation and ensuring that families felt welcome when coming to the site. Most sites sent notification of enrollment letters and after-visit summaries to the patient’s primary cardiologist and made a point to greet participants at non-study clinical appointments. Less frequently used strategies were certificates from study staff to the participant recognizing their participation and cards acknowledging special occasions (e.g., birthdays and/or holidays). One minor difference was that the majority (5/6) of higher adherence sites used calendars and trackers for both site staff and families to keep track of visits compared to 2/4 lower adherence sites.

Protocol adherence (visits/echocardiograms, Holters and quarterly calls)

There were no noteworthy differences in strategies between higher and lower protocol adherence sites. Most sites scheduled activities early in the window, scheduled future appointments before the subject left the current visit and provided travel funds when a family could not otherwise participate.

Study medication adherence strategies

All sites reported providing written and verbal instructions on administering the study medication and how to incorporate it into their daily routine. Additional methods included pre-paid mailers for bottle return (59%), incentives for bottle return (47%) and recommendations for mobile applications to provide reminders to take the study medication (41%).

Consistent with recruitment and retention strategies, sites with better study medication adherence reported using a greater number of strategies compared to lower adherence sites. These sites employed reinforcement of medication adherence and bottle return during study visits and calls with families, individualized calendars, pill crushers and text message reminders. Higher adherence sites encouraged use of reminder alarms and electronic pill boxes to manage medication schedules whereas lower adherence sites used these strategies less often. All higher adherence sites provided information cards about the study medication with emergency contact information, while all lower adherence sites reported seldom or never using these.

Marfan Foundation Partnership

The Marfan Foundation extensively promoted and advocated for trial participation by providing information via the internet, newsletters, conferences and highlighting patient stories. A Foundation social worker and nurse were available to families to answer questions and concerns about the trial and participated in study committee discussions. The Foundation provided regular updates on enrollment through multiple social media outlets. It actively educated physicians by disseminating targeted outreach materials for pediatric cardiologists and attending professional meetings for cardiologists and geneticists. It provided financial support for echocardiograms, genetic testing and travel stipends. They assisted 89 families over the course of the study by providing travel assistance (47 hotel stays, ground transportation 53 times, and 8 Mercy Medical airlifts), paying for 167 clinical care echocardiograms for families without insurance coverage and 22 “other insurance” expenses). They also paid for genetic testing for 13 subjects to confirm a diagnosis of Marfan syndrome for trial eligibility.

Discussion

Effective recruitment, retention and adherence are critical to obtain a proper sample size, ensure adequate power and effectively answer key scientific questions. High rates of protocol adherence enhance the validity of the trial results.¹⁰ Despite a sparse participant pool, the Marfan Trial successfully recruited the proposed sample size. Overall completion rates for in-person study visits/echocardiograms, Holters and quarterly safety calls were excellent, with the clear majority completed within pre-specified study windows. We speculate that the expertise, infrastructure and experience of the Pediatric Heart Network [www.pediatricheartnetwork.org]¹¹ contributed to this success. Trials conducted within one of the National Heart, Lung and Blood Institute’s funded networks were found to predict recruitment success.¹² These networks have the luxury of stable, yet flexible infrastructures (a dedicated Data Coordinating Center, highly experienced clinical sites and dedicated NIH Program staff) and can devote their efforts and resources to launching and monitoring studies, rather than vetting research sites and creating new clinical study processes for each study.

The Pediatric Heart Network’s Recruitment and Retention Plan and associated checklist highlight strategies for site recruitment and retention and formed the basis for the site questionnaire. The questionnaire results in Table 6 elucidate many of the strategies that sites employed, but it is not possible to say which strategies contributed most to successful recruitment and retention. However, five key lessons can be highlighted from this study related to differences in compliance by age and race, site differences, the value of collaborative learning, and partnerships with patient advocacy groups.

Younger age and better compliance:

By design, the Marfan Trial enrolled individuals from 6 months to 25 years of age to optimize early treatment to prevent additional aortic root enlargement. Younger age was significantly associated with higher adherence with in-person visits, Holter monitoring and quarterly call completion, as well as medication bottles returned. Young adults had the lowest adherence for in-person visits, especially for visits within window. They were also less likely to wear Holter monitors, participate in quarterly calls and return medication bottles. Not unexpectedly, they were typically enrolled in college or working and likely had more competing priorities, whereas the youngest participants had parents to support and reinforce adherence to the study regimen. These findings mirror clinical observations of adherence difficulties during adolescence and the transition to early adult care.^13–14

After analyzing data from this study, the Network recognized that recruitment, retention and adherence strategies for the harder to reach adolescents and young adults should be targeted. Different communication strategies, such as text messaging, apps for medication reminders, email, social media and incentives like i-Tunes or Amazon gift cards may be more effective with adolescents and young adults. Therefore, in another complex drug study of adolescents with a rare form of congenital heart disease, the Network sought feedback from parents and participants that resulted in a participant compensation plan and protocol that would enhance participation and reduce study burden. Participants also received an iPod to help provide visit reminders and to assist with medication adherence.¹⁵ The trial successfully enrolled 400 subjects in 23 months.¹⁶

Racial differences and compliance:

Family history of Marfan syndrome, female sex and history of neurodevelopmental disorder were found to have significant associations in only one aspect of protocol adherence (visit, Holter and quarterly call completion, respectively), and may be chance findings. Race was associated with multiple measures of adherence. Race and ethnicity are known to affect enrollment and retention/attrition rates in clinical trials,^17–18 and reasons for the impact of race on protocol and medication adherence should be pursued in future research, along with targeted interventions.

Site differences impacted success:

Clinical research site was associated with all measures of adherence. Variability between sites is common in multi-site clinical research. A study conducted by Shore et al.,¹⁹ showed that site management strategies were associated with medication adherence in patients being treated with dabigatran for atrial fibrillation. In our study, while all sites reported using successful strategies, such as building relationships with participants and referring providers, higher adherence sites were more likely to utilize a greater number of strategies. Examining variability between sites can reveal opportunities for learning and process improvement. A well thought out and comprehensive recruitment, retention and adherence plan is more likely to reach a broader audience and meet the diverse needs of the study population.

The variability of recruitment, retention and adherence strategies observed in this study were likely influenced by factors such as clinical research staff, infrastructure, resources and clinical volume. Most higher adherence sites had study staff continuity over the course of the trial, which was conducted over seven years, whereas most lower adherence sites reported at least one occurrence of turnover. A stable study team can more easily foster relationships with study participants, particularly in a study requiring a three-year commitment.^20–21There was also considerable variability in the number of staff working on the study at a given time at sites. Higher adherence sites tended to have greater staffing resources.

The extra strategies that higher adherence sites reported using may in part be due to greater staffing resources, allowing for additions such as greeting patients at non-study visits to the hospital and sending holiday and special occasion cards. It may also be that having more personnel resulted in more ideas for improving protocol adherence and permitted time to cultivate collaborations with referral physicians outside of the site’s typical catchment area. However, many of the strategies reported could easily be incorporated into visits, such as soliciting feedback from participants, strategizing with participants about medication reminders, and providing simple tools like calendars, trackers or emergency contact cards. The Network now develops specific recruitment/retention plans and materials prior to each study’s launch to support sites, particularly those with fewer resources.

Collaborative learning fosters success:

Medication adherence is essential to obtain meaningful trial results and for optimal clinical care. Yet, a Cochrane Collaboration²² of interventions for enhancing medication adherence, found that people typically take 50% of their prescribed doses of self-administered medication. Following an interim analysis, the Network realized the bottle return rate was not what it needed to be in order to adequately determine medication adherence. Thus, a conference call was used to discuss bottle return rates, and sites with high return rates shared strategies employed. Bottle return rates increased modestly, yet significantly, following the call, which highlights the importance of a collaborative learning approach. The Network continues to dedicate portions of study calls to sharing strategies and lessons learned among sites. The Network also applies pertinent lessons learned from other trials to better inform future trials.²³ Trials should include adherence plans at study design and conduct early assessments of adherence to identify potential problems and allow for the opportunity to implement improvement strategies.

Patient advocacy partnering:

A novel recruitment and retention strategy was the partnership with The Marfan Foundation. This helped the patient population to be properly informed, educated and confident about the trial. This, along with the work done at the sites, resulted in 97% of the screened and eligible patients consenting to enrollment and randomization.² The Foundation also helped to tackle some of the recruitment and retention challenges which eliminated many barriers to participation. Given the challenges of conducting clinical research in a rare disease population, partnering with advocacy groups was regarded as key to the success of the trial and has been successfully replicated in another trial launched by the Network.¹⁵

Limitations

There was a notable time lapse and staff turnover between completion of the trial and distribution of the site questionnaire to individual sites. Therefore, the collected data are subject to recall bias and some personnel were not available to complete the site questionnaire. A few sites did not return the site questionnaire, and there may be differences between sites who responded to the questionnaire and those who did not. Also, it is not possible to quantify the effects of strategies employed based on the nature of the site questionnaire and study design. There was wide variation in the percentage of bottles returned by site. We could only assess adherence based on the bottles that were returned. While adherence was good in those bottles returned, we could not assess adherence in those not returned; therefore, our estimate of overall medication adherence may be overly optimistic.

Conclusions

Overall, adherence was excellent for this trial conducted within a NIH-funded clinical trial network, however differences between sites were substantial. Higher adherence sites employed a greater number of strategies, many that any site could easily use. These sites also collaborated frequently with referral providers, utilized the full spectrum of resources offered by the patient advocacy organization, and had less staff turnover. Adolescent, young adult and African American participants were less adherent, and strategies specifically targeted to them were needed. These lessons learned could help improve recruitment, retention and adherence in future trials involving rare conditions.

Table 3:

Factors Associated with Holter Completion Percentage Significant covariates only (p-value<0.05)

Category	N	Mean Adherence Overall (%)	Bivariate Modelling p-value	Multivariable Modelling p-value	Mean Adherence In-window (%)	Bivariate Modelling p-value	Multivariable Modelling p-value
Total	593	94			84
Age			0.002	0.001		0.01	0.001
6 months to < 8 years old	212	96			87
8 to < 13 years old	156	95			85
13 to 18 years old	136	91			81
>18 years old	89	91			79
Sex			0.04	0.04		0.46
Male	357	93			84
Female	236	96			85
Race*			0.06			0.01	0.03
White	511	94			84
African American	46	89			76
Asian	16	99			85
American Indian or Alaskan Native	3	100			100
Native Hawaiian or Pacific Islander	3	100			100
More than one race	14	99			90
Clinical Research Site†			0.004	0.002		<0.001	<0.001
	6 – 56	88 – 100			57 – 95

^*due to small sample sizes, the p-value reflects only the comparison of Whites to African Americans

^†Range of sample sizes and mean adherences across the 21 sites

APPENDIX

Appendix of investigators

Pediatric Heart Network Investigators

In addition to the authors, the following investigators participated in the Pediatric Heart Network Marfan Trial:

National Heart, Lung, and Blood Institute:

Gail Pearson, Mario Stylianou, Victoria Pemberton

Network Chair:

Lynn Mahony, University of Texas Southwestern Medical Center

Data Coordinating Center:

New England Research Institutes, Lynn Sleeper* (PI), Sharon Tennstedt* (PI), Steven Colan*, Gloria Klein*, Lin Guey*, Lisa Wruck*, Thomas Travison*, Shan Chen*, Eric Gerstenberger*, Tanya Olesker*, David F. Teitel*

Core Clinical Site Investigators:

Boston Children’s Hospital, Boston, MA: Jane Newburger (PI), Ronald V. Lacro (Study Co-chair), Martha King*, Carolyn Dunbar-Masterson, Jill Handisides, Andrea Posa*, Quincy Nang*, Cara Hass; Children’s Hospital of New York, NY: Daphne Hsu (PI)*, Wyman Lai (PI)*, William Hellenbrand*, Beth Printz*, Mary J. Roman, Richard Devereux, Rosalind Korsin, Greysi Sherwood*; Children’s Hospital of Philadelphia, PA: Victoria Vetter (PI), Stephen Paridon, Marie Gleason, Reed Pyeritz; Nicole Mirarchi*, Sandra DiLullo*, Agbenu Ejembi, Ruth Morgan*, Tonia Morrison; Cincinnati Children’s Medical Center, OH: D. Woodrow Benson* (PI), William Border*, James Cnota, Haleh Heydarian, Jeanne James*, Michelle Hamstra, Kathryn Hogan*, Lois Bogenschutz*, Mary Pat Benham, Teresa Barnard - deceased; Duke University, NC: Page A. W. Anderson (PI) - deceased, Jennifer S. Li (PI), Stephanie Burns Wechsler, Amanda Cook*, Charles Sang, Wesley Covitz*, Mingfen Xu, Lori Jo Sutton, Kari Crawford*, Summer Roberts*, Deborah Palmer; Medical University of South Carolina: J. Philip Saul* (PI), Andrew Atz, Geoffrey Forbus, Teresa Atz, Patricia Infinger, Aparna Choudhury*; Primary Children’s Hospital and the University of Utah, Salt Lake City, UT: LuAnn Minich (PI), Richard Williams, Angela Yetman*, Marian Shearrow, Michelle Robinson*, June Porter*; Hospital for Sick Children, Toronto, Canada: Brian McCrindle (PI), Timothy J. Bradley*, Jennifer Russell, Jack Colman, Elizabeth Radojewski*, Svetlana Khaikin*, Nancy Slater*; Johns Hopkins University School of Medicine, MD: Harry C. Dietz (Study Co-chair), William J. Ravekes, Mary Rykiel, Elisabeth Sparks - deceased, Gretchen MacCarrick, Jennifer Leadroot*

Auxiliary Site Investigators:

Washington University School of Medicine, St. Louis, MO: Charles Canter (PI), Angela Sharkey*, Alan Braverman, Cheryl Rainey; Texas Children’s Hospital Houston, TX: John L. Jefferies*, Timothy Slesnick*, Aimee Liou (PI), Hugo Martinez*, Andres Menesses*, Tunu Tenende*; Stanford University Medical Center, Stanford, CA: David Liang (PI), Elisabeth Merkel; Ghent University Hospital, Ghent, Belgium: Bart Loeys*, Julie De Backer (PI), Jan Maarten Cobben, Thierry Sluysmans, Anne De Paepe, Sylvia De Nobele; Mount Sinai Hospital, New York, NY: Bruce Gelb (PI), Shubhika Srivastava, Tejani Mendiz-Ramdeen*, Constance Weismann, Emily Lawrence, Stephanie Chin, Helen Ko, Jen Le Yau; Children’s Hospital of Pittsburgh of UPMC, Pittsburgh, PA: Steven Webber* (PI), Stacey Drant (PI), Jane Luce, Kevin Stiegler*; Vanderbilt University, Nashville, TN: Larry Markham (PI)*, Cheryl Kinnard*, Cheri Stewart, Sue Sommers, Carol Madison; Ann & Robert H. Lurie Children’s Hospital, Chicago, IL: Luciana Young* (PI), Megan Domenico*, Kathryn Waitzman*, Carla Lozano*; Children’s Hospital and Clinics of Minnesota, St. Paul, MN: Mary Ella Pierpont (PI), Charles Baker, Erin Zielinski*, Heidi Vander Velden, Alison Overman*; Seattle Children’s Hospital, Seattle, WA: Mark Lewin (PI), Aaron Olson, Amy Payne; Cedars-Sinai Medical Center, Los Angeles, CA: David Rimoin (PI)- deceased, Mitchel Pariani*, Robert Siegel (PI), Asim Rafique*; Rady Children’s Hospital, UCSD, San Diego, CA: Paul Grossfeld, Arlene Smith, Terri McLees-Palinkas*

Echocardiography Core Laboratory:

Boston Children’s Hospital: Steven D. Colan (Director), Elif Seda Selamet Tierney*, Jami Levine, Shari Trevey, Marga Rivera

Protocol Review Committee:

Michael Artman, Chair; Erle Austin, H. Scott Baldwin, Daniel Bernstein, Timothy Feltes, Julie Johnson, Thomas Klitzner, Jeffrey Krischer, G. Paul Matherne, Kenneth G. Zahka

Data and Safety Monitoring Board:

John Kugler, Chair; David J. Driscoll, Mark Galantowicz, Sally A. Hunsberger, Thomas J. Knight, Holly Taylor

* No longer at the institution listed