Clinical Trial Details
— Status: Completed
Administrative data
NCT number |
NCT05338216 |
Other study ID # |
UL1TR002544 |
Secondary ID |
UL1TR002544 |
Status |
Completed |
Phase |
N/A
|
First received |
|
Last updated |
|
Start date |
December 19, 2022 |
Est. completion date |
January 31, 2024 |
Study information
Verified date |
March 2024 |
Source |
Northeastern University |
Contact |
n/a |
Is FDA regulated |
No |
Health authority |
|
Study type |
Interventional
|
Clinical Trial Summary
People with post-stroke aphasia (PWA) suffer from anomia, a condition where they know what
they want to say but cannot retrieve the words. For PWA, word retrieval changes
moment-to-moment, leading to diminished motivation to participate in conversations and
disengagement from social interactions. In the real world, anomia variability and severity
are compounded by contextual factors of communication exchanges (noise, dual-tasking).
Ecological momentary assessment (EMA) involves in-situ measurement of a behavior over time
during everyday life. EMA has promise for capturing real-world anomia, yet EMA methods have
not been tested in PWA. Therefore, the aims of this pilot study are to (1) determine the
relative feasibility of two types of smartwatch-delivered EMA (traditional-EMA and micro-EMA)
in PWA and (2) determine the extent to which patient-specific factors relate to feasibility.
Twenty PWA will be recruited and randomly assigned to either traditional-EMA or micro-EMA
conditions. To target in-situ anomia, PWA will complete 36 picture-naming trials/day for
three weeks, delivered either as a single trial 36 times per day (micro-EMA) or in four sets
of nine trials/set per day (traditional-EMA). Due to the "at-a-glance" single trial delivery
of micro-EMA, the investigators hypothesize that PWA in the micro-EMA condition will
demonstrate better protocol adherence than PWA in the traditional-EMA condition. Older age,
more severe cognitive-linguistic deficits, and greater discomfort with technology will be
related to poorer compliance, lower completion, greater perceived burden, and lower
intelligibility of naming audio recordings. This bench-to-bedside research will begin a
translational path to implement EMA/micro-EMA into routine assessment of aphasia.
Description:
BACKGROUND AND TRANSLATIONAL SIGNIFICANCE Background Approximately 15 million individuals
worldwide, including 795,000 Americans, suffer a new stroke each year. Aphasia, a disorder
characterized by receptive and/or expressive language deficits, affects approximately 30% of
stroke survivors and is one of the most devastating post-stroke conditions. Aphasia is a
heterogenous disorder, yet anomia (i.e., impaired word retrieval) is a persistent, ubiquitous
impairment present in all people with aphasia (PWA). Neurotypical individuals occasionally
experience anomia during "tip-of-the-tongue" moments: frustrating instances when someone
knows what they want to say but the word itself eludes them. For the two million Americans
currently living with aphasia, word retrieval difficulties are far more severe and pervade
all communication attempts. Anomia severity ranges between PWA, from mild difficulties such
as delayed retrieval of low frequency words (e.g., amulet) to severe difficulty retrieving
the names of even common objects (e.g., bed). Yet, regardless of severity, a hallmark of
post-stroke anomia is inconsistent word retrieval across attempts for a given word. For PWA,
the ability to access words can change day-to-day and even minute-to-minute, resulting in
diminished motivation to participate in conversations, disengagement from social
interactions, and reduced quality of life.
Per psycholinguistic models of lexical access, naming is a two-step process that involves
retrieval of the target's meaning from the semantic system and word form (speech sounds) from
the phonological system. Naming failures in PWA result in errors that are related to a target
word in meaning (e.g., "carrot" for bean), word form (e.g., "fean" for bean), both meaning
and form (e.g., "green" for bean), or neither (e.g., "pencil" for bean; "I know it but I
can't say it" for bean) due to insufficient activation of the target meaning/form
representations or overactivation of incorrect representations. Due to this noisy system,
lexical access is often delayed in PWA, even when the correct word is eventually retrieved.
Naming accuracy, response times, and error types provide meaningful information regarding the
locus of naming impairment and a patient's recovery capacity. The investigators found that
the number of phonological and unrelated errors produced during early post-stroke stages
predicted longitudinal changes in naming and global language skills in left hemisphere stroke
survivors. Naming errors and the integrity of underlying semantic and phonological systems
can also guide language treatment planning. As such, accurate assessment of naming is
critical for aphasia care.
In clinical practice, speech-language pathologists (SLPs) typically assess naming deficits
through a single administration of a test in a quiet clinic room in a 1:1 interaction between
the SLP and patient. Yet, given the variability in word retrieval, a one-time assessment
cannot accurately capture the true nature and extent of a patient's anomia. Moreover, unlike
a controlled, contrived clinical environment, word retrieval in the real world usually occurs
while an individual performs other daily tasks, often in distracting, noisy environments and
with many other people present. The added cognitive load of multi-tasking and suppressing
external distractors adds to an already noisy lexical system, thereby increasing the chance
of delayed or failed word retrieval. Traditional SLP assessment cannot capture the realities
of real-world anomia, but the investigators argue that ecological momentary assessment (EMA)
can.
EMA is a widely used method in health research to capture longitudinal self-reported
behaviors or states of interest (e.g., physical activity, depression, diet) in daily life. In
traditional EMA, an electronic device prompts a participant to answer a set of questions
(usually via button press) many times a day over time. Answering several questions at a time
is time intensive, and thus, a key disadvantage of traditional EMA is interruption burden. In
contrast, micro-interaction-based EMA (µEMA) reduces each prompt to a single question that
can be answered in ~3-5s, allowing for many more questions interspersed over time. The
investigators' pilot studies suggest that even with an interruption rate eight times more
than traditional EMA, µEMA has higher response rates and lower perceived burden than
traditional EMA. Furthermore, µEMA yields temporally dense in-situ measures, making it an
ideal tool for capturing the variability of anomia in the daily lives of PWA. One caveat is
that an anomia µEMA/EMA protocol would necessitate recording verbal responses, yet
traditional EMA and µEMA protocols to date have only required nonverbal responses via a
button press or screen tap. The feasibility of capturing high-quality audio responses in a
µEMA or traditional EMA protocol is unknown.
Moreover, despite the increasing use of technology-based assessments and therapy platforms
for aphasia, neither electronically delivered EMA nor µEMA has been tested in PWA. In fact,
based on a review from 2020, only seven studies have used EMA to study stroke recovery and
none targeted language impairments. A central question is whether stroke survivors with
aphasia can adhere to electronically delivered traditional EMA or µEMA anomia protocols.
Prior research regarding tablet and computer-based platforms for aphasia indicates that PWA
with better visuospatial attention and executive functions are better able to learn how to
navigate a novel iPad application. Evidence suggests that SLP support is associated with
greater adherence to a computerized therapy program for PWA and that patients' perceived
engagement in the computer program is related to opportunities for practice, their ability to
use the technology, and their motivation. An important step towards broader implementation of
EMA and/or µEMA in aphasia is determining PWA's adherence to such protocols and which
patient-specific factors are associated with protocol adherence.
Thus, as a future step toward broader-scope research goals, in this pilot, the investigators
will (1) determine the relative feasibility of two smartwatch-delivered overt naming EMA
protocols in PWA and (2) determine the extent to which patient-specific factors (age,
cognitive-linguistic deficit severity, and comfort with technology) relate to feasibility.
All participants will be scheduled to complete the same number of naming trials over the
course of the experiment, but some PWA will be randomly assigned to a traditional EMA
protocol (involving four prompts per day with nine naming trials per prompt) whereas other
PWA will be assigned to a µEMA protocol (involving a single naming trial delivered at a time
36 times per day). Feasibility will be measured in three ways: in terms of (1) patients'
ability to adhere to the EMA/µEMA protocol, (2) patients' perceived burden of completing the
protocol, and (3) the capacity to capture recorded audio of naming attempts. The first two
feasibility measures are similar to those used in the investigators' prior work, whereas the
final measure is unique to this study since this pilot constitutes the first attempt at using
EMA/µEMA to objectively measure verbal responses in stroke survivors
Innovation Despite the small scope of this pilot study, several innovative features comprise
this research. This pilot will be the first study to evaluate the feasibility of
electronically delivered EMA protocols in stroke survivors with aphasia. The comparison of
traditional EMA and µEMA protocols in a clinical population is also novel. This work will be
critical in determining which delivery method works best for PWA who differ from
previously-studied neurotypical samples in meaningful ways. In addition, EMA research
typically utilizes self-report measures to catalogue mental or cognitive states in daily
life. In contrast, this pilot will attempt to objectively measure the cognitive-linguistic
skill of naming through audio recordings of naming attempts in patients' daily lives. Such a
paradigm has not been attempted before and thus, it is simultaneously original and risky with
the potential for high reward. If it is feasible to obtain intelligible naming responses in
an EMA/µEMA paradigm, this pilot study will yield unprecedented knowledge about PWAs'
real-world experiences with anomia. If it is not feasible to obtain high-quality audio
recordings from many participants, this pilot will still yield valuable knowledge regarding
patients' ability to follow EMA/µEMA protocols and their perceived burden of EMA/µEMA.
Translational Significance This research is at the T1: From Bench to Bedside stage, given
that the primary goal of this proposal is to evaluate the viability of EMA/µEMA as a novel
assessment tool of language impairments in PWA. Translational potential is intrinsic to
EMA/µEMA methods. The field of SLP lacks objective measures of communication deficits in
real-life situations and the novelty and innovation of this work lies in working to remedy
this problem. Findings from this study will inform a future NIH R01 application focused on
verifying the criterion validity of EMA/µEMA in capturing anomia (in terms of accuracy,
response rate, and naming error types) and determining the impact EMA/µEMA-based anomia has
on participation and quality of life in PWA compared to traditional anomia assessment. If
successful, these methods can be expanded to measure other language domains (e.g.,
single-word to phrase-level auditory or reading comprehension) with personalized tasks
tailored to patients' needs. Information gleaned from this study and the NIH R01 work also
will be used to develop in-situ, "just-in-time" ecological momentary interventions (EMI) for
anomia in which smartwatch-delivered prompts can intervene each time a patient experiences an
anomic moment. The NIH R01 proposal, EMA/µEMA work into expanded language domains, and
development of "just-in-time" EMI for anomia constitute three future directions of this pilot
that also fall at the T1: From Bench to Bedside stage of the translational continuum.
Therefore, in the future, the investigators will conduct research at the T2: From Bedside to
Practice stage that will involve partnering with practicing SLPs and other healthcare
professionals (e.g., physical and occupational therapists, physiatrists) at area
rehabilitation hospitals or university-based clinics (e.g., Northeastern University Speech,
Language, and Hearing Center) to conduct EMA/µEMA and EMI clinical trials in PWA. This work
will yield vital information about the efficacy of these methods in real clinical practice.
Successful completion of T1 and T2 phase studies also will pave the way for work at the level
of T3: From Clinical Practice to Widespread Clinical Practice and Care Delivery. Research at
the T3 stage is realistic for this line of work given that smart technology costs will likely
continue to drop, making smartwatches and phones more accessible to more people, and the
variety and flexibility of digital mobile health applications will continue to rise,
resulting in potential widespread clinical use of EMA/µEMA and/or EMI protocols. Thus, this
research will begin a translational path that ideally ends in implementation of EMA/µEMA and
EMI into routine clinical care of aphasia that is reimbursable by insurance providers.
SPECIFIC AIMS The long-term goal of this line of research is to be able to reliably measure
real-world language deficits in people with aphasia (PWA) with EMA/µEMA and to develop
effective "just-in-time" interventions for PWA to improve their ease of everyday
communication. As the first step towards this goal, the main objective of this proposal is to
determine the relative feasibility of smartwatch-delivered micro-interaction ecological
momentary assessment (µEMA) and traditional EMA as potential evaluation methods of aphasia.
The secondary objective of this proposal is to determine which patient-specific factors (age,
cognitive-linguistic deficit severity, and comfort with technology) are related to
feasibility metrics.
Twenty PWA will participate in this pilot study, randomly assigned to the µEMA condition (n =
10) or traditional EMA condition (n = 10). In both the µEMA and traditional EMA protocols,
the smartwatch will vibrate to alert the participant to a prompt to complete either a single
picture naming trial (in the µEMA condition) or a set of picture naming trials (in the
traditional EMA condition). After the vibration, the participant will see a "Ready?" screen
in which they must press a YES button for the picture to appear, and then they will attempt
to name the picture aloud. Participants in both conditions will be scheduled to complete 36
overt picture naming trials per day for three weeks. The only difference between conditions
will be the trial delivery schedule: PWA in the µEMA condition will complete a single naming
trial at a time, 36 times per day whereas PWA in the traditional EMA condition will receive
four prompts per day to complete a set of nine picture naming trials per prompt. The two-fold
rationale for comparing µEMA and traditional EMA protocols is that the investigators will be
able to ensure the results reflect the viability of µEMA or traditional EMA specifically and
not just the novelty of smartwatch use, and this approach will provide preliminary data
regarding which method works best for which PWA.
As previously referenced, investigators will index feasibility in three ways. First,
investigators will measure patients' adherence to the µEMA or traditional EMA schedule via
two objective measures: compliance and completion. Because smartwatch µEMA/EMA has not been
tested in PWA, investigators do not know PWAs' capacity for adhering to the protocol by
pressing a button or providing a verbal response. Therefore, investigators will disentangle
feasibility of button press responses to smartwatch-delivered µEMA/EMA from the feasibility
of obtaining intelligible audio from the overt naming protocol. To do so, the YES/NO button
tap responses to the "Ready?" screen will be used to calculate compliance. Similar to the
investigators' prior work, compliance will be determined by considering the total number of
YES button responses over the total number of scheduled prompts, including trials missed due
to the watch being uncharged or turned off. On the other hand, completion will reflect PWAs'
adherence to all smartwatch protocol steps, including pressing the YES button, seeing the
picture(s), and attempting to name the picture(s) aloud. Therefore, completion will be
calculated by considering the total number of naming attempts (indexed by the number of audio
clips with an audible voice recording, disregarding intelligibility) over the total number of
prompts delivered to the watch, excluding trials missed due to the watch being uncharged or
off. Second, investigators will obtain patient-reported measures of perceived burden measured
via Likert-scale questions on weekly surveys administered during weeks 2-4. Third, SLP
graduate student RAs will code each picture naming audio clip as being either completely
intelligible, partially intelligible, or completely unintelligible; this categorical speech
intelligibility rating will provide a gross measure of feasibility of obtaining audio
recordings of naming attempts in µEMA/EMA in PWA. With these data, investigators will address
the following aims:
Aim #1: To compare feasibility measures (compliance, completion, perceived burden, and speech
intelligibility) between the µEMA and traditional EMA conditions. Hypothesis: Based on the
investigators' prior work, investigators predict PWA in the µEMA condition will demonstrate
higher compliance and completion and report lower perceived burden than PWA in the
traditional EMA condition. Investigators also predict that speech intelligibility will be
greater in the µEMA condition than the traditional EMA condition because situational factors
that disrupt recordings in a given moment (e.g., background noise, distance between the
participant's mouth and the smartwatch) will be more likely to disrupt several trials in a
row (as in the traditional EMA condition) than trials that are interspersed throughout the
day (as in the µEMA condition).
Aim #2: To determine the extent to which patient-specific factors (age, cognitive-linguistic
deficit severity, and comfort with technology) relate to feasibility measures. Hypothesis:
Controlling for study condition, investigators predict that older age, more severe aphasia,
worse non-linguistic cognitive skills (visuospatial and executive control abilities), and
greater discomfort with technology will be associated with poorer compliance, lower
completion, higher perceived burden, and poorer intelligibility of audio recordings during
naming attempts across the entire group (n = 20 PWA).
RESEARCH PROCEDURES Approach Participants - Twenty adults with aphasia (18-89 years old) will
complete the proposed protocol. Inclusion criteria will be: (1) current/pre-stroke English
proficiency, (2) normal/corrected-to-normal vision and hearing, (3) medical stability, (4)
history of left hemisphere stroke at least six months prior to enrollment, and (5) presence
of aphasia as determined by consideration of scores on language assessments (e.g., Quick
Aphasia Battery, discourse tasks) and the study team's clinical judgment (spearheaded by Dr.
Meier). The reason for this two-pronged approach for determining aphasia presence is that
widely-used comprehensive aphasia batteries are not reliably sensitive to mild aphasia,
particularly for detecting less severe anomia. The exclusion criterion will be a history of
neurological disease affecting the brain besides stroke.
Methodology - This study is for a longitudinal, controlled observational study. PWA will be
randomly assigned to either a µEMA or traditional EMA condition.
Week 1: Participants will complete two, two-hour sessions: (1) an initial intake and
evaluation session followed by (2) an EMA/µEMA protocol training session. Week 1, session 1
tests will include: (1) the Quick Aphasia Battery (QAB) to measure overall aphasia severity,
(2) the Pattern Comparison Processing Speed Test and Flanker Inhibitory Control and Attention
Test from the NIH Cognition Toolbox to measure non-linguistic visual attention and executive
functions, respectively, (3) brief, standardized picture description and story elicitation
tasks to index anomia in discourse, and (4) an overt object naming test to index word
retrieval via a standard SLP assessment format (i.e., single administration in a quiet 1:1
interaction). The naming test will include normed real photos of the 260 items from the
Snodgrass and Vanderwart stimuli set. PWA will also participate in an intake interview and
complete a questionnaire about their comfort with technology, adapted for aphasia (simple
language, short phrases) with responses scaled from 1 = strongly disagree to 7 = strongly
agree. Due to their aphasia, PWA will require more training on using the smartwatch than
neurotypical samples from prior µEMA/EMA studies. Thus, week 1, session 2 will focus on
training activities, including:
1. Smartwatch Orientation: PWA will be oriented to the smartwatch (interface, charging,
etc.) through simple verbal and written instructions and visual demonstrations.
2. EMA Task Training: PWA will be instructed on the specific EMA/µEMA naming task,
including instructions on how to tap on the screen and the optimal approach for
providing verbal responses (i.e., mouth close to the watch). Then, PWA will complete an
18-item traditional EMA-style naming task probe (i.e., several trials back-to-back on
the smartwatch) in a quiet room in The Aphasia Network (TAN) Lab at Northeastern
University. During this probe, SLP graduate student research assistants (RAs) will
provide cueing and feedback so that PWA successfully learn how to operate the smartwatch
and respond to prompts.
3. µEMA or EMA Protocol Simulation: PWA will complete a "real-world" µEMA or EMA task probe
in a distracting environment (i.e., coffee shop or university bookstore on
Northeastern's campus). For this final activity, over a 45-minute period, 18 smartwatch
prompts will be given, either randomly interspersed with one naming trial per prompt
(for PWA in the µEMA condition) or split into two sets of 9 trials per prompt (for PWA
in the EMA condition). This final activity will mirror the week 2-4 µEMA/EMA protocols.
Weeks 2-4: PWA will complete either the µEMA or EMA protocol, depending on randomization.
µEMA and EMA protocols will be delivered via a Fossil Sport Android smartwatch (or
equivalent) provided by the study team. In each condition, the smartwatch will vibrate to
alert PWA to a single naming trial (µEMA condition) or naming trial set (traditional EMA
condition). After the vibration alert, PWA will see a screen that says "Ready to name a
picture?" (µEMA condition) or "Ready to name some pictures?" (traditional EMA condition). If
the participant taps YES, a picture will appear, and the device will begin recording audio.
PWA will have up to five seconds to provide an oral response, after which a "Thanks!" screen
will appear. The five-second response window was selected based on research showing that PWA
correctly name pictures within 3.5 seconds, on average, with longer response latencies for
incorrect responses. If a participant fails to respond to a vibration prompt or pushes the NO
button to the "Ready?" screen, they will be re-prompted five minutes later via one more
vibration alert.
During weeks 2-4, each participant will attempt to name 108 unique objects derived from the
260-item Snodgrass and Vanderwart photoset. The subset of pictures presented to each
participant will be based on their naming evaluation (during week 1, session 1). When
possible, half of the pictures will be items the PWA named correctly during the evaluation,
and the other half will be incorrectly named pictures. This approach will allow us to gauge
naming variability over time of items PWA are able versus unable to name during traditional
SLP assessment, providing an important window into anomia. To mitigate practice effects, the
individual pictures (µEMA condition) or picture sets (traditional EMA condition) will be
presented randomly without replacement until all 108 pictures have been presented, and then
the picture cycle will restart. In both conditions, 36 naming trials will be scheduled per
day, resulting in a total of 756 scheduled trials for each participant across the experiment.
In the µEMA condition, single-naming trial prompts will appear at random intervals from 10am
to 8pm. In the EMA condition, PWA will receive four prompts per day to complete a set of nine
picture naming trials per prompt; one prompt will be scheduled in every 2.5-hour block
between 10am and 8pm.
Each night, participants will place the watch on a charger. Each morning, a smartphone
provided by the study team will beep to alert the participant to take the smartwatch off the
charger. The smartphone also will push the data collected by the smartwatch to the study
team's server. At the end of each week, PWA will complete a 16-item survey about their
µEMA/EMA experiences during the prior week. Surveys will be administered via Qualtrics and as
needed, will be facilitated by SLP graduate student RAs via video conferencing software.
Week 5: In a final session at TAN Lab, PWA will return the smartwatch, complete the 260-item
Snodgrass and Vanderwart picture naming test, and participate in a study exit interview.
Family, friends, or care providers involved in assisting the PWA with any aspects of the
protocol will also be asked to complete an exit interview.
Statistical Analyses - Aim #1: Investigators will code each trial in the µEMA or EMA time
series to index compliance, completion, and speech intelligibility. Every scheduled trial (n
= 756 total trials) will be coded as 0 for a non-response/NO button response or 1 for a YES
button response, reflecting compliance. Every delivered trial with a naming attempt captured
on audio (n trials will vary by person) will be coded as 0 for incomplete or 1 for complete,
reflecting completion. Every audio clip (n trials will vary by person, depending on the
number of naming attempts) will be coded as 0 for completely or partially unintelligible and
1 for completely intelligible, reflecting speech intelligibility. For each of these three
measures, investigators will run a logistic mixed effects model with one of the feasibility
measures (i.e., compliance, completion, or speech intelligibility) as the dependent variable,
condition (µEMA or EMA) as the independent variable, and random effects of participant and
trial. For perceived burden, investigators will collect responses on three questions/survey
for three weeks. Given the small amount of data, survey responses will be plotted to
visualize trends in increasing/decreasing burden rather than compared statistically.
Aim #2: Investigators will conduct a series of partial correlations between patient-specific
factors and feasibility measures (compliance, completion, speech intelligibility, and
perceived burden), controlling for study condition. Patient-specific factors will include
age, aphasia severity (derived from the QAB), non-linguistic cognitive skills (averaged
performance on the two NIH Toolbox tasks), and a summary score derived from the comfort with
technology survey. Here, compliance will be the proportion of YES button responses divided by
total scheduled trials (YESbutton/756) for each participant. Completion will be the
proportion of naming attempts divided by total delivered trials (n naming attempts/n trials
delivered). Speech intelligibility will be the proportion of completely intelligible
responses divided by the total number of naming attempts (n completely intelligible trials/n
naming attempts). Perceived burden will be calculated as the sum of all burden question
Likert responses. Multiple comparison correction will be done at a false discovery rate of p
< 0.05.
Sample size/power: For the Aim #1 logistic mixed effects models, two groups of 10 PWA per
group will achieve 80% power to detect an odds ratio of 1.20 in a design of 756 repeated
measures with a AR(1) covariance structure and when the correlation between observations of
the same subject is 0.600 and alpha is 0.05. If the proportion of completed trials for the
EMA group is 0.500, a statistically significant result will occur if the µEMA group
proportion is 0.545 or greater. In Intille et al, the mean difference in proportions between
the µEMA and EMA conditions was >0.20 for compliance and completion, which is much higher
than the needed 0.045 difference. This suggests investigators will have ample power to detect
significant differences between the two conditions. For Aim #2, to achieve 80% power with 20
participants and alpha at 0.05, Pearson correlation coefficients will need to be 0.59 or
greater. Due to the small but feasible sample of this pilot, investigators may not achieve
such effects, but these data will be critical for power calculations for the NIH R01.
Further, investigators will have a wealth of information to describe qualitative trends in
patient-specific factors, even if significance for Aim #2 is not achieved.