Inter-Dose Variability of Immunosuppressant Medication Among Adolescent Heart Transplants During Video Directly Observed Therapy

Abstract

Background:

Medication adherence and timing of doses remains a challenge for adolescent patients following heart transplant. Mobile technology and adherence promotion efforts offer new avenues for direct observation of medication adherence and timeliness of medication-taking behavior. The study explores posttransplant medication maintenance, highlighting the importance of consistent dose timing with tacrolimus. The use of directly observed therapy (DOT) via a mobile health app was examined as a method for real-time medication monitoring, offered a platform for patients to upload videos of themselves taking medication for review by transplant team members.

Methods:

The study examined a single-group design involving 10 adolescent heart transplant recipients over a 12-week DOT intervention, assessing both medication adherence and post-intervention outcomes.

Results:

Results from multilevel regression models analyzing inter-dose timing and deviation from 12-h intervals revealed significant variability among patients and a correlation between increased deviation from the 12-h dose interval and both older patients and those with greater perceived barriers to medication adherence.

Conclusions:

Findings suggested a link between deviation from recommended dose timing and poorer posttransplant health outcomes. Findings underscored the potential of DOT and mobile health to examine timeliness of medication adherence, to directly observe medication-taking behavior, and association with posttransplant health outcomes.

1 |. Introduction

Adolescent patients often struggle with posttransplant medication adherence, and promoting better medication-taking behavior remains a significant challenge for transplant teams [1]. Posttransplant medication maintenance therapy for heart recipients typically consists of a three-drug regimen including a corticosteroid, a calcineurin inhibitor such as tacrolimus, and a purine synthesis antagonist or an mTOR inhibitor. Internationally, the combination of tacrolimus and mycophenolate mofetil is the most commonly prescribed medication regimen for this population [2].

Consistency in tacrolimus dose timing as much as general adherence to a daily medication plan is paramount to successful outcomes. Stable tacrolimus levels maximize successful outcomes while minimizing adverse effects. The typically prescribed dose of tacrolimus should be taken in 12-h intervals to achieve recommended stable blood trough concentrations to support therapeutic efficacy while limiting toxicity [3–5]. The pharmacokinetic profile of tacrolimus can be affected by the presence of food (i.e., fasted vs. fed conditions) [4], the composition of the diet (i.e., high fat vs. high carbohydrate) [4, 5], the formulation of tacrolimus taken (e.g., differing C_min between IR-Tac and ER-Tac and longer T_max with LCPT) [4–6], patient age [7], and the race of the patient (i.e., lower C_max and AUC_0–inf in Black patients compared to White patients) [5].

Low tacrolimus levels increase the risk of rejection and graft loss, while high levels increase the possibility of infection and renal complications [8–10]. Dependent on a patient’s clearance value, a single missed dose of tacrolimus can significantly impact trough concentration levels, with up to a 49% decrease in simulations of kidney transplant patients with high drug clearance [11]. These missed or excessive doses of tacrolimus can cause incremental damage that may go unnoticed until complications occur. Regular blood level monitoring of tacrolimus is necessary to ensure that drug levels remain within the therapeutic range and to adjust dosing as needed to maintain efficacy and minimize toxicity. Despite the use of blood level monitoring, real-time monitoring of medication adherence remains a challenge for both patients and transplant care teams.

Exact monitoring of patient adherence and measurement of medication-taking behavior is inherently difficult for transplant teams and researchers [1]. Gaining direct observation of medication-taking is even more difficult and likely logistically impossible for any length of time barring when patients receive inpatient care. Many studies and healthcare providers use indirect or proxy measures for obtaining data on medication-taking behavior [12]. Patient self-reports, pill counts, and pharmacy refill records each offer inexpensive and easily obtained data on medication adherence yet are notoriously unreliable and do not offer direct information on medication-taking behavior. Electronic drug monitoring using devices have been suggested, such as the medication event monitoring system (MEMS) cap. Despite its apparent utility, these devices often suffer from data loss, patients forgetting to return the device for data download, hinder patients using medication organizers (e.g., pillboxes), and other data management issues. Even with these issues, the central disconnect remains when equating removal of a pill cap with actually ingesting the medication. More recently, pharmaceutical innovations have resulted in digital medication wherein a small sensor is included in a pill capsule with the medication. When ingested, the sensor admits a signal detected by an external detector device worn by the patient (on a lanyard or included in an adhesive patch worn near the patient’s stomach) [13, 14]. Studies have found the use of digital medication with pediatric patients to be challenging including difficulties with external detector, problems with the user interface, and low acceptability with pediatric patients. Adolescent patients especially may experience greater challenges and less engagement with the digital medication monitoring [14].

Directly observed therapy (DOT) offers the opportunity for promote medication adherence among patients and offers real time medication monitoring. A mobile health (mHealth) application developed by Scene Mobile Health Inc. allows patients to upload videos of themselves taking their medication to a platform immediately following recording for asynchronous review by transplant team members [15, 16]. The platform and mobile application additionally allow for interpersonal engagement and rapport building between patients and transplant staff. Additionally, the DOT app provides a supportive feedback system for the patient, behavioral cues for medication-taking, and actively encourages each the development of the patient’s organizational skills, planning, and conscientiousness around the behavior. With the use of DOT, providers can identify patterns of nonadherence and swiftly intervene when necessary. Instead of waiting until lab work is collected and analyzed, healthcare teams can immediately contact patients who have submitted videos of tacrolimus doses taken outside of 12-h intervals.

A 12-week DOT pilot intervention study was conducted with a small sample of adolescent heart transplant (HT) recipients [16]. The DOT intervention with these patients experience demonstrated high levels of acceptability and engagement early in the intervention [17]. Patients completing the 12-week program submitted 90.1% of possible videos which demonstrated excellent medication adherence and later clinically significant improvements in long-term MLVI [15]. An analysis of in-app communication and messages between patients and nurses found DOT as promoting patient engagement which included a varied nature of interpersonal support and patient education [18]. The current study aimed to explore inter-dose timing and deviation from 12-h intervals with this sample. We examined the utility of directly observed inter-dose timing and deviation from 12-h dose intervals to predict each patients’ observed deviation from 12-h dosing interval and each patient characteristics, dose characteristics, medication adherence from blood test values, perceived barriers to medication adherence, and long-term posttransplant outcomes. Findings from the study have important implications including importance of tacrolimus dose timing, the impact of deviation from 12-h intervals, and the clinical utility of directly observing medication adherence using a mHealth platform.

2 |. Materials and Methods

2.1 |. Study Design

A single-group design for a 12-week DOT intervention involved 10 transplant recipients ages ranging from 10 to 21 and their parents [15, 16]. Baseline quality of life and adherence were assessed with patients and their parents using a pre-intervention survey. Feasibility and acceptability were assessed via a phone interview at 3 weeks [17]. At 12 weeks, subjects completed a post-intervention survey and an exit interview.

2.2 |. Procedure

Participants were recruited from a large southeastern pediatric HT program in the United States. Remote telephonic and scheduled evaluations during routine outpatient clinic visits were used in recruitment efforts, patient consent, and patient enrollment. Post-consent, researchers reached out to patients and their parents with pre-intervention surveys to assess baseline measures. Scene Health Inc. personnel provided patients with downloading and installing the DOT application on their mobile devices. Patients and their parents were contacted by researchers upon completion of the 12-week program for debriefing and post-intervention interviews. Enrollment, training, measurement, and staff responsibilities during the DOT intervention were published previously [16]. Per the protocol, if a patient did not submit a video demonstrating medication adherence for seven consecutive days despite attempted contact and re-engagement of the patient, the patient was considered non-participatory. The University of Florida Institutional Review Board approved the study (#IRB202000695).

2.3 |. Sample

Eligibility requirements included patients who received a HT at least 6 months before enrollment, were between the ages of 10 and 21 years, have a history of medication adherence difficulties, and be determined by the transplant team as medically stable. Patients with medication level variability index (MLVI) [19] values ≥2.0 or recommended by the transplant team were identified and recruited.

2.4 |. Measures

2.4.1 |. Inter-Dose Timing, Intervals and Deviation

Patients submitted videos as evidence of medication-taking behavior according to their medication plan, which in the current study was twice per day tacrolimus. Videos and their exact time of submission were used in calculations of time between evidence of administration of doses of tacrolimus. Time between observed medication taking was calculated for each dose for each patient, ideally 12-h from previous dose. For analyses, deviation from the 12-h dosing interval was used. Deviation was calculated as the absolute value of the number of hours since the previous dose observed minus 12 h. For example, doses taken either 9 or 15 h after the previous dose would be consider equally a 3-h deviation in from the 12-h dosing interval.

2.4.2 |. Measurement of Adherence and Long-Term Health Outcomes

Adverse medical outcomes for all patients were recorded from transplant center electronic health records. Each instance of acute rejection and hospitalization considered were those during the intervention period and 6 months post-intervention. Rejection was defined as biopsy-proven acute rejection confirmed during inpatient care.

2.4.3 |. Medication Level Variability Index

Medication Level Variability Index values were calculated as the standard deviation (SD) of tacrolimus blood levels per patient [12, 19], such that less consistency in medication-taking behavior leads to greater variability in blood levels and higher MLVI values. In a large multisite study of pediatric liver transplant recipients, MLVI scores were predictive of acute rejection [19]. At least three blood level tests were used to calculate the tacrolimus levels for each patient [19, 20]. MLVI values were determined as part of the screening process 6 months prior to enrollment, during the DOT intervention, and again in 6-month follow-up periods. As an outcome in the current study, MLVI was calculated as the intra-patient SD value using blood levels taken during the DOT intervention and 6-month post-intervention periods.

2.4.4 |. Barriers to Adherence

Adolescent self-reported perceived barriers to medication adherence and parents’ proxy-reported barriers to their child taking medication were assessed using the Adolescent Medication Barriers Scale (AMBS) and the Parent Medication Barriers Scale (PMBS), respectively [21]. AMBS and PMBS each contain three subscales which include Disease Frustration/Adolescent Issues, Ingestion Issues, and Regimen Adaptation/Cognitive. Reliability, construct validity, and predictive validity of previous research about adolescent transplant recipients and their parents have been demonstrated in the overall measure of each version [22].

2.5 |. Statistical Analysis

Descriptive statistics detailed the inter-dose timing and deviation from 12-h intervals. Bivariate analyses examined differences in dosing intervals across patients and completion of the DOT intervention. Next, multilevel regression models (MLM) were conducted to estimate the variation among doses submitted across the 12-week program and among all patients. MLM estimates variance in an outcome when data are nested at multiple levels (i.e., doses nested within patients). Firstly, a random intercept only model was conducted to estimate the variance at each level of data (i.e., Level 1 at doses, Level 2 at patients) using intra-class correlation (ICC). ICC is the proportion of variance of the outcome at each level of the data. Next, dose characteristics at Level 1 was used to predict deviation from the 12-h dose interval, including timing of the dose (morning compared to evening), weekday, and number of days spent in the 12-week program. Level 2 predictors included patient characteristics such as age, gender, age at transplant, minority status, private versus public insurance, status at transplant, current parent age, and each patient self-reported and parental-reported AMBS and PMBS scores. Estimates of slope and significance tests (t-tests with 95% confidence intervals) were calculated. Lastly, MLVI values and adverse outcomes such as hospitalization due to rejection and mortality were included in nonparametric Mann–Whitney U-test as outcomes of deviation from 12-h inter-dose timing intervals.

3 |. Results

3.1 |. Sample Demographics

Overall, 10 of 14 patients approached for participation initiated the 12-week DOT program (Table 1) and ranged in age from 12 to 21 years. Eight patients completed all 12 weeks of the program. Most of those initiating DOT were female, had private insurance, and had a pretransplant diagnosis of congenital heart disease, cardiomyopathy, or graft rejection requiring a retransplantation. Among these patients who initiated the 12-week DOT program, 90.1% of doses were observed directly via patient-submitted videos and verified asynchronously by transplant nursing staff (see Table 2).

TABLE 1 |.

Patient and parent characteristics, n = 10.

Patient characteristics
Patient age, years, M ± SD	16.60 ± 2.95
Median	15
Range (min.–max.)	12–21
Gender, female, n (%)	6 (60.0%)
Race, minority, n (%)	2 (20.0%)
Parent age, years, M ± SD	39.29 ± 3.35
Type of insurance at transplant, n (%)
Medicaid	4 (40.0%)
Private	6 (60.0%)
Patient hospitalized at transplant, n (%)	8 (80.0%)
Patient in ICU at transplant, n (%)	4 (40.0%)
Coronary allograft vasculopathy (CAV), n (%)
None	2 (20.0%)
Mild	6 (60.0%)
Moderate to severe	2 (20.0%)
Proportion of submitted dose videos	76.7 ± 29.9%
Median	86.6%
Range (min.–max.)	17%−100%
MLVI values at DOT enrollment, M ± SD	2.86 ± 1.83
Median	2.19
Range (min.–max.)	0.95–6.17
MLVI >2.0, n (%)	6 (60.0%)
MLVI values during and 6 months after DOT, M ± SD	2.08 ± 2.02
Median	2.02
Range (min.–max.)	0.70–3.79
MLVI >2.0, n (%)	5 (50.0%)
Inter-dose interval time, M ± SD	13.42 ± 7.33
Median	12.02
Range (min.–max.)	0.97–167.51
Inter-dose deviation from 12-h, M ± SD	2.11 ± 7.16
Median	0.38
Range (min.–max.)	0.00–155.51

TABLE 2 |.

Predictors of deviation from 12-h dose intervals.

Predictor	Estimate	t-test	95%CI LL	95%CI UL
Submission level, n = 1342
Dose (evening)	−0.24	0.65	−0.96	0.48
Weekend dose (yes)	0.37	0.92	−0.42	1.17
Weeks in DOT	0.20	3.73***	0.10	0.31
Patient level, n = 10
Patient age	0.76	2.21*	0.10	1.44
Gender (male)	1.05	0.42	−3.73	5.82
Minority (yes)	−0.95	0.31	−6.84	4.91
Private insurance (yes)	1.26	0.51	−3.49	6.04
Hospitalized at transplant (yes)	−1.14	0.37	−7.11	4.79
ICU at transplant (yes)	1.59	0.65	−3.12	6.28
VAD at transplant (yes)	2.83	0.98	−2.77	8.42
Intubated at transplant (yes)	−0.11	0.03	−7.99	7.75
Parent age	0.08	0.36	−0.34	0.50
Adolescent Medication Barrier Scale, AMBS, post-intervention
Total scale score	0.17	2.66*	0.05	0.30
Disease Frustration/Adolescent Issues score	0.21	2.68*	0.06	0.37
Regimen Adaptation/Cognitive Issues score	0.54	1.85	−0.02	1.10
Ingestion Issues score	−0.08	0.24	−0.69	0.54
Parental Medication Barrier Scale, PMBS, post-intervention
Total scale score	0.05	0.45	−0.17	0.28
Disease Frustration/Adolescent Issues score	−0.14	0.56	−0.62	0.34
Regimen Adaptation/Cognitive Issues score	0.13	0.72	−0.22	0.48
Ingestion Issues score	−0.01	0.01	−1.25	1.25
Reminder score	0.47	1.13	−0.32	1.26

^*p < 0.05,

^**p < 0.01,

^***p < 0.001.

3.2 |. Inter-Dose Timing and Variability

From the 10 patients, a total of 1342 videos were submitted and counted as evidence of adherence to immunosuppressant medication. Mean inter-dose interval was 13.42 ± 7.33 h with a median of 12.02 h. Deviation from 12 h schedule per dose was calculated. Mean deviation from inter-dose 12-h interval was 2.11 ± 7.16 h with a median value of 0.38 h (22.8 min). For these 10 patients, including those two patients eventually leaving the program after only 3 weeks, the range of deviation from 12-h intervals ranges from 0.01 to 155.51 h. Figure 1 provides a visual of the mean and standard deviations for each patient, for all morning and evening doses, across their participation in the 12-week DOT intervention.

FIGURE 1 |.

Ten participants and dose times for tacrolimus across a 84-day DOT intervention (individual patients’ dose times shown, 1a through 1j). Doses presented on 24-h clocks with morning (red) and evening (blue) doses. Mean morning and evening dose time per patient shown by lines in clock. Mean and standard deviations for dose times given as box and whisker plot on clock edge. Day within an 84-day intervention shown outward from clock center. Patients shown in (1c, 1j) did not complete 84-day intervention.

Across all patients, inter-dose interval (F_9,1332 = 19.36, p < 0.001, η² = 0.116) and the inter-dose deviation from 12-h periods (F_9,1332 = 22.41, p < 0.001, η² = 0.132) each significantly differed among patients. The patients completing only 3 weeks of the DOT program demonstrated significantly greater inter-dose timing deviation (t₁₃₃₀ = 11.64, p < 0.001, d = 1.27).

3.3 |. Multilevel Modeling Predicting Inter-Dose Deviation From 12 h

Multilevel regression models was then used to examine predictors of inter-dose deviation from 12-h periods for tacrolimus. A random intercept only model indicated that 22.26% of the variance in inter-dose deviation values is among the different patients (ICC = 22.26). There were no significant differences in inter-dose deviation times between the each the morning and evening doses and those doses taken on the weekend compared to the weekday. Weeks in the program significantly predicted deviation from 12-h interval (β = 0.20, t₁₃₃₂ = 3.727, p < 0.001) with each week spent in the DOT program associated 0.20 h (12 min) greater deviation from patients’ 12-h dose interval. Similarly, patient age significantly predicted (β = 0.76, t_7.96 = 2.20, p = 0.050) an increased difference of 0.76 h (45.6 min) in patients’ 12-h dose interval for every year the patient was older. No other demographic or medical variable significantly predicted deviation from 12-h dosing interval.

3.4 |. Medication Adherence, Barriers, and Inter-Dose Deviation From 12 h

Changes in MLVI values calculated from the 6 months prior and compared to MLVI values in the 6 months following DOT were tested for association with deviation from 12-h intervals among patient doses. Changes in MLVI from pre- to post-intervention were positively correlated with inter-dose deviation from 12-h intervals within the 10 patients (r = 0.474, p = 0.244, r² = 0.225). Though not statistically significant with such a small sample, the correlation represented a moderate effect with patient deviation from 12-h intervals explaining 22.5% of changes in MLVI from pre- to post-intervention. Post-intervention AMBS scores significantly predicted deviation from 12-h intervals with AMBS total (β = 0.17, t₁₃₃₂ = 2.660, p = 0.04) and AMBS Disease Frustration/Adolescent Issues (β = 0.21, t₁₃₃₂ = 2.680, p = 0.04) each predicting greater deviation from 12-h intervals.

3.5 |. Post-Intervention Health Outcomes and Inter-Dose Deviation From 12 h

Association between deviation from 12-h intervals among patient doses and adverse outcomes such as hospitalization due to rejection and mortality, though not statistically significant due to a small sample. Those seven patients with no adverse outcomes in the 6 months following participation in DOT has a mean inter-dose deviation interval of 1.84 ± 1.72 h compared to mean inter-dose deviation of 6.67 ± 5.40 h three patients experiencing an adverse health outcome (Mann–Whitney U-test z = 1.25, p = 0.210). Despite lack of statistical significance, the effect size was large (Hedge’s g = 1.41) demonstrating greater dose timing variability for those patients who later experienced rejection and hospitalization following participation in the DOT intervention.

4 |. Discussion

The study sought examine inter-dose timing and interval for a small sample of adolescent HT recipients and its association with dose characteristics, patient types, medication adherence factors, and posttransplant outcomes. Inter-dose timing variability was found to differ significantly among patients in the sample and across their time in the 12-week DOT program. Results identified nearly a quarter of the variability in dose timing was between patients, similar to findings from other longitudinal adherence studies where significant variation in adherence was among patients and less so over time [20, 23, 24].

Multilevel regression models models identified time in DOT and patient characteristics as predictive of patients’ deviation from 12-h dose intervals. Older patients and those reporting greater perceived barriers to medication adherence were each predictive of greater deviation and less consistency with the 12-h dose interval. These findings mirror a wealth of prior research with adolescent patients often experiencing patterns of lower medication adherence emerges during developmental period [10, 12, 25]. This period for patients often includes assuming of the responsibility for completing health behaviors, as adolescents increasingly manage their health in combination with familial and social strains [10, 26]. Deviation from 12-h intervals did not vary by evening or morning doses and by the day of the week.

Deviation from inter-dose 12-h interval was associated with other biomedical indicators of nonadherence and poorer posttransplant health outcomes. Despite the small sample size and lack of statistical significance, the improvements in MLVI values over the course of the intervention and post-intervention period was strongly associated with less deviation from inter-dose 12-h intervals. Similarly, the association between dose timing variability and experiences of rejection and hospitalization was very large.

The current study exploring inter-dose timing and deviation within a sample of adolescent HT recipients had several limitations that must be considered. The small sample size from a singular transplant center reduces the reliability and generalizability of the results. Patient attrition from the intervention was associated with higher MLVI values and older patients [15], similar to that of other mHealth studies with adolescent transplant recipients [14]. Importantly, patients already experiencing greater nonadherence are difficult to recruit into research [27, 28] likely due to inconsistent clinical attendance. Current research efforts are being made to recruit a larger sample of adolescent HT recipients to increase generalizability, reliability of statistics estimates, and produce a more robust examination of inter-dose timing and DOT [29].

The findings from the current study lend support to DOT as a powerful tool to assess inter-dose timeliness even during an intervention aiming to promote medication adherence. While numerous other methods of measuring and assessing medication adherence have distinct challenges and disadvantages [20, 30], DOT through a mobile app allowed for twice-daily check-ins to offer direct assessment of inconsistency with medications and a potential direct indicator of future risk of nonadherence. Additionally, and perhaps most importantly, the consistency of timing of doses was found to be predictive of a biomedical marker of medication adherence and posttransplant outcomes. DOT and inter-dose interval might offer a novel method for the detection of medication adherence in the shorter term, in addition to number of videos submitted demonstrating medication-taking behaviors, allowing for more immediate action taken by transplant care coordination teams.

Abbreviations:

AMBS

Adolescent Medication Barriers Scale

DOT

video directly observed therapy

EHR

patient electronic medical record

HT

heart transplant

ICC

intra-class correlation

MEMS

medication event monitoring system

mHealth

mobile health

MLM

multilevel regression models

MLVI

medication-level variability index

mTOR

mammalian target of rapamycin

PMBS

Parent Medication Barriers Scale

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.