Digital biomarkers: Redefining clinical outcomes and the concept of meaningful change

Abstract

MCID (minimal clinically important difference) is a patient‐centered concept used in clinical research that represents the smallest change that someone living with Alzheimer's disease would identify as important. There are several challenges associated with the universal application of this construct. Alzheimer's disease progresses differently for each individual, complicating the establishment of a universal standard that accounts for individual‐level issues. Alzheimer's disease is also a gradual and evolving disorder, and what is perceived as clinically meaningful can vary significantly at early and late disease stages. People living with Alzheimer's disease and caregivers may have differing perspectives on the benefits of treatment outcomes, making it more challenging to establish an appropriate MCID. Moreover, Alzheimer's trials rely on a variety of tests to evaluate cognitive and functional impairments. However, these tests often lack sensitivity to early‐stage changes and are affected by variability in rater rankings. Digital biomarkers and advanced health technologies have emerged as a hot topic in modern medicine. They offer a promising approach for detecting real‐time, objective clinical differences and improving patient outcomes by enabling continuous monitoring, individualized assessments, and leveraging artificial intelligence learning for complex analytical predictions. However, while these advancements hold great potential, they also raise important considerations around standardization, accuracy, and integration into current clinical frameworks. As new technologies are introduced alongside evolving regulatory frameworks, the primary focus must remain on outcomes that truly matter to people living with Alzheimer's disease and their caregivers, ensuring that the principle of clinical meaningfulness is not lost.

Highlights

• Minimal clinically important difference (MCID) represents the smallest change in a patient's condition that would be considered meaningful, but defining this for Alzheimer's disease is challenging due to its heterogeneity.

• The perception of what is clinically meaningful may differ at the individual level, at different disease stages within the same individual, and between patient and caregiver.

• Traditional tests used as endpoints in Alzheimer's trials lack the sensitivity to detect subtle changes and are limited by range restrictions, making them less effective for accurately capturing treatment efficacy.

• Digital biomarkers and artificial intelligence (AI)‐driven health technologies may offer the potential to enhance the detection of clinically meaningful changes by providing continuous, objective monitoring and advanced analytics for individualized patient assessments.

• Both the United States Food and Drug Administration (FDA) and European Medicines Agency (EMA) are playing pivotal roles in advancing the use of digital health technologies, facilitating the evolution of regulatory frameworks to ensure these innovations are effectively integrated into clinical research and practice.

1. NUANCES OF MCID IN ALZHEIMER'S DISEASE RESEARCH

MCID (minimal clinically important difference) is a patient‐centered concept used in clinical research that evaluates whether a change in a patient's health condition is meaningful or beneficial from their perspective. In clinical trials of Alzheimer's disease (AD), this typically refers to a positive treatment outcome that is quantified as an improvement in cognitive and functional abilities, reflecting a meaningful improvement in their quality of life. ¹ Conceptually, it represents the smallest change that someone living with AD would identify as important.

However, there are several challenges attached to the universal application of this construct. First, the context of applying an MCID in a population that includes both employed individuals and those who have been retired for several years. There are numerous cases of people seeking medical help due to changes in their cognition that have affected their job performance. Such individuals may notice very small reductions in their cognitive function that impact their ability to perform effectively in their roles. In contrast, retired individuals might experience marked changes in their cognitive skills without noticing. Given that MCID calculation typically takes place at the group level, this fails to take account of individual‐level issues. ²

A further issue is soliciting the view of individuals who may lack insight into their condition, either due to frank anosognosia or an inability to articulate their status. This lack of insight can complicate the assessment of what constitutes meaningful change or benefit for the individual. In this case, patient and caregiver perspectives on treatment outcomes may differ on what is considered beneficial for the individual. Caregivers often have a comprehensive understanding of the patient's condition and may be able to recognize cognitive and functional changes that the patient does not perceive. ³ This divergence can complicate the establishment of an appropriate MCID as an objective measure that is centered on patients’ needs.

Moreover, AD is a progressive disorder that evolves gradually and differently across individuals. ⁴ Neuropathological changes build up over decades, first with no noticeable symptoms, but followed by the earliest signs of cognitive difficulties, which later impact the ability to function independently. Patients may even experience cognitive, behavioral, or functional changes at different rates, some exhibit rapid progression, while others may not progress at all within a certain time frame. Mild cognitive impairment (MCI) and AD can present very differently in the earliest stages of the disease. This variability is captured explicitly in our taxonomies of MCI. Perhaps unsurprisingly, individuals with episodic memory difficulties account for the majority of MCI cases. ⁵ However, deficits in the domains of attention, vision, language, and executive function can often be the presenting feature of MCI. ⁶ Oddly, the distinctions between MCI subtypes are blurred as individuals transition to a diagnosis of AD. Pivotal clinical trials for therapies like donanemab and lecanemab ⁷ , ⁸ ; have aimed to capture cognitive and functional change using multidomain instruments, consistent with regulatory guidance, such as the Clinical Dementia Rating (CDR) scale and the integrated Alzheimer's Disease Rating Scale (iADRS). However, concerns remain regarding their sensitivity to early and subtle cognitive decline, as both instruments may be less equipped to detect nuanced changes in attention, visuospatial processing, or executive function. This highlights the opportunity for digital assessments to complement existing tools by offering higher‐resolution insights into cognitive domains that are often underrepresented.

Gaps in key cognitive data and the complex and heterogeneous progression of AD further complicate the establishment of a universal standard for MCID. With individuals varying substantially regarding their key presenting features and variable rates of progression, establishing what is considered meaningful for a patient at the early disease stages is currently extremely challenging. This heterogeneity of disease progression makes it especially complex to assess treatment efficacy. Digital health technologies (DHTs) present a unique opportunity to capture cognitive as well as functional changes. However, for DHTs to serve as reliable assessment tools, their ability to detect changes that meet MCID criteria must be comprehensive. Indeed, a structured approach is necessary to determine whether DHT‐derived metrics fully or partially translate into meaningful improvements in patients’ daily lives. While MCID is a widely recognized concept, other methodological approaches, such as minimal important change (MIC) offer a complementary perspective in defining clinically meaningful treatment effects. MIC refers to the smallest change in a measure that patients would perceive as important; however, unlike MCID, it does not require a specific (or predefined) threshold. Instead, it accounts for individual variability and allows for a more nuanced understanding of gradual symptom progression. This distinction is crucial, as cognitive decline is a continuous rather than binary process, making MIC particularly useful for tracking subtle functional improvements that may not reach group‐level MCID thresholds.

In clinical trials, researchers should account for the variability in cognitive and functional impairments that are present in individuals living with AD. Traditional pen‐and‐paper tests, such as the Alzheimer's Disease Assessment Scale – Cognitive (ADAS‐Cog), when used as endpoints pose a dual challenge; on the one hand, they may not be sensitive enough to detect subtle cognitive changes early in the disease; on the other hand, they can detect statistically significant treatment effects that fail to meet MCID thresholds for many patients. This issue has been highlighted in recent anti‐amyloid trials, where observed cognitive benefits were statistically robust but clinically marginal. The core limitation of current assessments, therefore, is not just their sensitivity, but also their ability to differentiate meaningful cognitive improvements from changes that are too small to be of real benefit. ⁹ These types of assessments, most of which were developed in the 1970s and 1980s, are also time‐consuming to administer, prone to variability in rater scoring, and limited by range restrictions (ceiling and floor effects), making them less effective for accurately tracking disease progression or capturing treatment efficacy. Doubts about statistically but perhaps not clinically significant effects on these measures induced the United States Food and Drug Administration (US FDA) to adopt a co‐primary approach, requiring positive treatment effects on both cognitive and functional tests. Cognitive assessments aim to quantify the core deficits caused by neurodegeneration, while functional evaluations capture how these deficits affect a patient's daily life. By integrating both dimensions, regulatory bodies seek to ensure that therapeutic benefits are evaluated not only through changes in test scores but also through improvements that are meaningful in real‐world contexts. It is evident that more sensitive and objective outcome measures are needed to effectively address the heterogeneity in AD and better assess the efficacy of new treatments.

2. HOW CAN DIGITAL BIOMARKERS FILL THE GAPS?

Digital biomarkers are defined as objective physiological and behavioral data collected and measured by digital technologies (wearables, smart devices, etc.). ¹⁰ Digital cognitive assessments have been in use for more than 30 years as computerized or smart‐device‐based versions of traditional pen‐and‐paper tests. In recent years, innovation in digital health technologies has seen a rise in new app‐based cognitive assessments. These newer measures assess the participant's cognition and employ artificial intelligence or machine learning to derive a clinical outcome prediction. ¹¹

Digital biomarkers have shown the potential to significantly contribute to the detection of meaningful change. By enabling frequent monitoring, they permit the collection of large amounts of data that could allow for the detection of subtle changes missed by traditional methods.

An important nuance of traditional endpoints is the comparison to a standardized norm, which may not be particularly meaningful at an individual level. Digital biomarkers offer the opportunity to collect individual baseline thresholds and additional longitudinal data points (in the clinic and through remote patient monitoring). ¹² This approach has the potential to detect minimal clinical differences in the context of a disease with high variability and insidious onset. This is particularly important for the early preclinical stages, where symptoms are known to be “silent” according to standard assessments, even in the presence of AD biomarker positivity.

By providing objective measures of physiological and behavioral parameters, digital biomarkers are less prone to range restrictions as well as subjective interpretations, which are common in psychometric evaluations. In addition, by leveraging AI and machine learning, large volumes of digital biomarker data can be analyzed to identify patterns, cluster groups, or trends that would otherwise be “invisible” to the eye of a healthcare professional. ¹³ Given that machine learning algorithms are trained on collected data, they can learn from expanding datasets and, as such, improve their accuracy, generalizability, and detection capabilities.

However, an important consideration when using digital biomarkers is that high‐resolution measurement does not automatically translate into clinical meaningfulness. While DHTs enable the detection of subtle cognitive and functional variations, not all of these changes will be relevant to disease progression or patient outcomes. A key risk is the potential for “over‐measurement,” detecting statistically significant variations that may or may not have an actual impact on a patient's quality of life. To avoid this pitfall, DHT validation must focus not only on the precision of measurement but also on ensuring that detected changes correlate with meaningful functional or cognitive benefits for the patient. As mentioned earlier, one of the key criticisms of current AD clinical trials is that traditional assessments often detect treatment effects that, while statistically significant, may not be meaningful from a patient or caregiver perspective. Digital biomarkers have the potential to address this limitation by capturing multi‐domain functional and cognitive changes that are directly relevant to patients' daily lives. For example, digital measures of speech fluency, reaction times, or real‐world executive function may offer more holistic patient‐centered endpoints compared to traditional episodic memory tests alone. By analyzing longitudinal real‐world data, DHTs can help define MCID thresholds that reflect meaningful improvements in function, rather than isolated statistically significant effects.

3. MCID AND MIC: CONSIDERATIONS FOR DEFINING MEANINGFUL CHANGE IN DIGITAL BIOMARKERS

As pointed out earlier, a key challenge in defining meaningful treatment effects is ensuring that statistically significant changes translate into real‐world patient benefits. MCID and MIC offer distinct but complementary approaches to evaluating clinical relevance. To ensure DHTs align with these clinical frameworks, rigorous validation strategies can be applied, such as (a) anchor‐based validation: correlating DHT‐derived changes with established clinical scales (e.g., ADAS‐Cog) to ensure that digital biomarkers reflect real‐world (and externally validated) cognitive improvements; (b) distribution‐based validation: evaluating whether detected changes exceed natural variability, ensuring they are not mere statistical noise; and (c) real‐world validation: assessing longitudinal patient‐reported outcomes to verify that DHT‐derived measures correspond with meaningful functional improvements.

Eventually, by combining data from multiple sensor outputs, such as motoric function, speech analysis, and brain activity, DHTs can offer a high‐resolution, individualized disease trajectory, ensuring that detected changes are both statistically and clinically relevant. ¹⁴ Also, to ensure that DHTs accurately capture meaningful clinical changes, a structured development and validation plan is required. This process might consider the following steps:

1. Data standardization and integration: DHTs must leverage multi‐domain inputs (e.g., motor function, speech analysis, reaction time, cognitive performance) and align these with traditional and validated clinical endpoints or functional scales. Standardizing digital outputs to ensure compatibility with traditional assessments is an important step in building clinically relevant DHT models.

2. Defining MCID thresholds: relying on longitudinal data, using real‐world and data from clinical trials can further support the definition of meaningful thresholds specific to different stages of AD progression.

3. Prospective validation: Once MCID thresholds are defined, a prospective validation study must confirm that these digital measures reliably track clinically meaningful changes over time. This is also of significant relevance for regulatory bodies.

A digital cognitive assessment with regulatory clearance and proven clinical performance would empower frontline healthcare providers to significantly reduce the time from initial concern to diagnosis and treatment by making informed evaluations. However, the adoption of digital biomarkers and AI/machine learning (ML)‐based assessments will be driven not only by demonstrating their clinical validity and utility, but also by integrating them into clinical decision support systems. To achieve widespread use, it will be crucial to overcome barriers such as technology literacy, data privacy protection, and the challenges of incorporating a new technology into long‐established practices with constrained workflows. Addressing this friction will be crucial to transform how neurodegenerative diseases are managed in frontline care.

4. PROGRESS IN DIGITAL BIOMARKER ADOPTION: REAL‐WORLD APPLICATIONS AND EVOLVING REGULATORY FRAMEWORKS

The integration of digital biomarkers into AD research and clinical practice has gained significant momentum in recent years. This has been driven by advances in DHTs and regulatory bodies' evolving perspectives on their utility. Novel digital measures must be rigorously evaluated against traditional clinical outcomes to ensure real‐world relevance. By integrating MCID‐driven validation frameworks, DHTs can provide more reliable indicators of disease progression and treatment response. The FDA has also demonstrated strong support for the use of DHTs in clinical drug development and even developed a core program to engage interested parties within this framework. ¹⁵

A major challenge in AD drug development is differentiating treatment effects that are statistically significant from those that provide tangible functional benefits. Traditional clinical assessments have struggled to make this distinction, as highlighted by recent anti‐amyloid trials. DHTs provide an alternative approach by incorporating continuous real‐world data, enabling researchers to define patient‐centered MCID thresholds rather than relying solely on group‐level statistical significance. Recent initiatives from the FDA and global consortia emphasize the need to align digital endpoints with meaningful patient outcomes, further supporting the integration of DHTs into AD clinical trials. This regulatory interest undoubtedly reflects significant advances in DHTs with the potential to revolutionize the tracking and assessment of neurodegenerative symptoms. ¹⁶

Public‐private consortia, such as the RADAR‐AD (remote assessment of disease and relapse—Alzheimer's disease), or IDEA‐FAST (identifying digital endpoints to assess fatigue, sleep, and activities of daily living [ADL] in neurodegenerative disorders and immune‐mediated inflammatory diseases), are two major examples of large‐scale research initiatives. ¹⁷ , ¹⁸ Both have focused on developing and validating digital biomarkers capable of tracking disease progression, cognitive decline, and other related symptoms in naturalistic settings, providing valuable data for clinical trials and improving patient care. While Alzheimer's research is still refining its digital endpoints, several other fields of medicine have successfully implemented MCID‐based frameworks for digital health technologies. The challenges of defining clinically meaningful treatment effects are not unique to AD. Other medical fields have extensively debated the application of MCID and MIC, and the challenges of defining clinically meaningful treatment effects extend indeed beyond Alzheimer's research. Studies evaluating postoperative recovery and physiotherapy have highlighted difficulties in defining fixed MCID thresholds, as patient‐perceived recovery can show significant variations. It is therefore important to keep in mind that MCID is context‐specific and needs to be determined, applied, and interpreted appropriately, as otherwise it can be misleading. ¹⁹

It is becoming clear that incorporating digital technologies offers many benefits (e.g., continuous monitoring), and this is of particular interest considering the limitations of non‐DHTs techniques such as classic pen‐and‐paper neuropsychological tests for cognition.

A key learning point from these consortia was that DHTs have significant potential for tracking disease progression. However, their successful adoption will require reliable digital biomarkers that are standardized, and that provide clinically meaningful data in a user‐friendly and secure manner. ²⁰ , ²¹ A further requirement is that they should create minimal friction when integrated into healthcare practices. This underscores the need for collaboration across sectors, including regulators, to ensure that these technologies can be effectively utilized in clinical drug development and real‐world settings.

As it getting more documented that DHTs offer continuous, objective, and scalable monitoring in the early stages of AD, and recognizing the immense opportunity for further research and clinical impact, the Digital Medicine Society (DiMe) has spearheaded an initiative to assess the characteristics of DHTs in AD and related dementias, resulting in a landscape review of digital health technologies for these groups. ²² This multi‐stakeholder landscape review has been now integrated into the Library of Digital Measurement Product, an open‐access resource hosted by DiMe for the clinical, research, and developer communities (DATAcc). ²³

The use of DHTs within the field of AD and related dementias is now at a pivotal intersection, bringing together technology start‐ups, academic centers of excellence, pharmaceutical companies, and regulatory bodies. Technology start‐ups are providing cutting‐edge innovations leveraging the power of ML, AI, and quantum computing to support clinical research, while the FDA is moving in parallel to provide a supportive framework. ²⁴ The FDA has shown openness regarding the use of digital biomarkers, and building on the 2021 draft guidance, has issued a final guidance document in 2023 providing recommendations on the use of DHTs for data acquisition in clinical investigations. ²⁵ This guidance emphasizes considerations for validation, data integrity, patient safety, and regulatory compliance, ensuring that DHTs are effectively and reliably integrated into clinical trials. In Europe, while there is not yet such specific guidance akin to the FDA's, the European Medicines Agency (EMA) has been actively engaging in the integration of DHTs in clinical trials, highlighting their impact on disease monitoring and personalized medicine, as outlined in the regulatory science strategy to 2025. ²⁶

In sum, multi‐stakeholder initiatives highlight the growing interest in digital biomarkers from DHTs for improving AD detection and treatment outcomes. While regulatory bodies are increasingly supportive of these innovations, widespread adoption will depend on continued efforts toward robust validation, standardization, and ensuring the data provided by digital biomarkers is clinically meaningful.

5. CONCLUSIONS

Conceptually, MCID represents the smallest change that someone living with AD would identify as important. However, the universal application of MCID faces challenges due to the lack of personalized focus in a highly variable disease and disease progression, and the subjectivity of traditional cognitive and functional endpoints. Digital biomarkers offer a promising alternative by providing objective, continuous, and individualized assessments, leveraging advanced technologies such as AI and machine learning for detecting subtle clinical differences and providing complex analytical predictions. As new technologies are introduced and regulatory frameworks adapt, it is essential to ensure that clinical meaningfulness remains central, enabling these innovations to effectively improve outcomes in AD research and patient care.

AUTHOR CONTRIBUTIONS

Maria Florencia Iulita, Emmanuel Streel, and John Harrison, developed and drafted the manuscript.

CONFLICT OF INTEREST STATEMENT

M.F.I. and E.S. are employees of Altoida, Inc. and may hold stock options in the company. J.H. is an employee and shareholder in Scottish Brain Sciences and a paid consultant to Altoida. The author disclosures are available in the Supplementary Information.

CONSENT STATEMENT

Consent from human subjects was not required for this Perspective article.

Supporting information

Supporting Information

Iulita MF, Streel E, Harrison J. Digital biomarkers: Redefining clinical outcomes and the concept of meaningful change. Alzheimer's Dement. 2025;11:e70114. 10.1002/trc2.70114