Irvine, CA – February 6, 2026 – Johnson & Johnson today announced 12-month pilot-phase data from the OMNY-AF study, evaluating the investigational OMNYPULSE Platform for the treatment of symptomatic paroxysmal atrial fibrillation (AFib), during the 31st Annual AF Symposium in Boston. Initial results for 12-month outcomes across the 30-patient pilot cohort show investigators achieved 100% acute procedural success with no procedure-associated adverse events, while 56.7% of cases were performed with zero fluoroscopy and 90% of patients achieved primary effectiveness at 12 months.i
OMNY-AF is a prospective, single-arm, multi-center clinical trial conducted across more than 40 sites in the U.S. and Australia.i The study pairs the OMNYPULSE Catheter, a 12 mm large-tip focal catheter featuring contact-force sensing and bipolar, biphasic pulse delivery with the TRUPULSE Generator. This integrated design combines precise mapping, controlled energy delivery and live feedback through the PF index on the CARTO 3 System.ii The OMNYPULSE Platform is not currently approved in any region of the world.
“The 12-month data provide encouraging early evidence on the OMNY-AF study with promising safety outcomes – no procedure-related adverse events or MRI-detected cerebral lesions – across eight centers in the pilot phase i. In my cases during the ongoing OMNY-AF trial, the seamless integration of advanced mapping, ultrasound, and PF Index with contact force were valuable for precise and efficient pulsed field energy delivery,” said Dinesh Sharma, M.D.1, Section Head of Cardiac Electrophysiology at the Naples Heart Institute, the study presenting author.
Alongside the OMNY-AF data, Johnson & Johnson is highlighting new findings related to the VARIPULSE Platform. Data presented by Andrea Natale, M.D.2, and simultaneously published in JACC Clinical Electrophysiology, by Moussa Mansour, M.D.3 examined the incidence of neurovascular events following the workflow enhancements and the introduction of an optimized irrigation flow rate. Notably, the platform sustained a low neurovascular event rate of 0.22% in 6,811 patients after implementation of both workflow enhancements and the updated irrigation rate.ii
Additional VARIPULSE Platform data presented at AF Symposium adds to the growing body of evidence underscoring the platform’s consistent and favorable safety profile across a range of clinical and real-world settings, including:
VARISURE Safety survey data presented by Christopher Porterfield, M.D.4: Early results from this physician survey on 850 procedures indicated low complication rates with a 1.9% rate of primary adverse events, a 0.2% incidence of neurovascular events and no reported cases of coronary spasm or death. Same-day discharge was achieved in 87.9% of patients.iii
REAL AF registry analysis presented by Mohammad-Ali Jazayeri, M.D.5: Results from the REAL AF registry showed excellent acute safety outcomes of the VARIPULSE Catheter, with a low overall acute safety event rate of 0.5% with no neurovascular events, high rates of same-day discharge and no observed differences in safety outcomes across AFib classifications.iv
Irrigation Flow Optimization research presented by Fengwei Zou, M.D.6: Preclinical data demonstrated parity between the 4 mL/min and 30 mL/min irrigation rates in microbubble generation, hemolysis and lesion depth when using the VARIPULSE Catheter, while confirming that higher irrigation significantly reduced electrode surface heating.v
“These data reinforce confidence in the consistency of safety outcomes observed across Johnson & Johnson’s electrophysiology portfolio. As a relatively new energy modality, pulse field ablation technologies should be individually evaluated for safety and reproducibility in atrial fibrillation ablation,” said Gregory Michaud, M.D., Chief Medical and Scientific Officer, Electrophysiology, MedTech, Johnson & Johnson. “As pulsed field ablation continues to evolve, rigorous evidence generation and transparent data sharing will be essential to advancing the science and enabling the next wave of innovation with this technology.”
Johnson & Johnson remains committed to evidence-driven innovation that advances patient care and informs clinical decision-making across its electrophysiology portfolio. These efforts are supported by the CARTO 3 System, the world’s leading cardiac mapping system7.
About The OMNY-AF Study The OMNY-AF study is a prospective, single-arm, multi-center study evaluating the clinical safety and effectiveness of the OMNYPULSE Catheter for the treatment of symptomatic paroxysmal AFib. Up to 440 enrolled subjects will undergo an ablation procedure with the OMNYPULSE Platform. The primary safety endpoint in the study is the occurrence of Primary Adverse Events within seven days of the ablation procedure. The primary effectiveness endpoint is freedom from documented (symptomatic and asymptomatic) atrial tachyarrhythmia episodes based on electrocardiographic data and additional failure modes during the effectiveness evaluation period over a 12-month period.
About the VARIPULSE Platform The VARIPULSE Platform is Johnson & Johnson MedTech’s Pulsed Field ablation system. The fully integrated platform includes the VARIPULSE Catheter, TRUPULSE Generator, and CARTO 3 Mapping System VARIPULSE Software. The Platform is now approved for use in the United States, Europe, Asia Pacific, Canada, and Latin America.
Cardiovascular Solutions from Johnson & Johnson MedTech Across Johnson & Johnson, we are tackling the world’s most complex and pervasive health challenges. Through a cardiovascular portfolio that provides healthcare professionals with advanced mapping and navigation, miniaturized tech, and precise ablation we are addressing conditions with significant unmet needs such as heart failure, coronary artery disease, stroke, and atrial fibrillation. We are the global leaders in heart recovery, circulatory restoration, and the treatment of heart rhythm disorders, as well as an emerging leader in neurovascular care, committed to taking on two of the leading causes of death worldwide in heart failure and stroke. For more, visit biosensewebster.com.
About Johnson & Johnson At Johnson & Johnson, we believe health is everything. Our strength in healthcare innovation empowers us to build a world where complex diseases are prevented, treated, and cured, where treatments are smarter and less invasive, and solutions are personal. Through our expertise in Innovative Medicine and MedTech, we are uniquely positioned to innovate across the full spectrum of healthcare solutions today to deliver the breakthroughs of tomorrow and profoundly impact health for humanity. Learn more about our MedTech sector’s global scale and deep expertise in surgery, orthopaedics, vision, and cardiovascular solutions at https://thenext.jnjmedtech.com. Follow us at @JNJMedTech and on LinkedIn.
Cautions Concerning Forward-Looking Statements This press release contains “forward-looking statements” as defined in the Private Securities Litigation Reform Act of 1995 related to Collaborative Outcomes Registry for Evidence in Ventricular Arrhythmias. The reader is cautioned not to rely on these forward-looking statements. These statements are based on current expectations of future events. If underlying assumptions prove inaccurate or known or unknown risks or uncertainties materialize, actual results could vary materially from the expectations and projections of Johnson & Johnson. Risks and uncertainties include, but are not limited to: competition, including technological advances, new products and patents attained by competitors; uncertainty of commercial success for new products; the ability of the company to successfully execute strategic plans; impact of business combinations and divestitures; challenges to patents; changes in behavior and spending patterns or financial distress of purchasers of health care products and services; and global health care reforms and trends toward health care cost containment. A further list and descriptions of these risks, uncertainties and other factors can be found in Johnson & Johnson’s most recent Annual Report on Form 10-K, including in the sections captioned “Cautionary Note Regarding Forward-Looking Statements” and “Item 1A. Risk Factors,” and in Johnson & Johnson’s subsequent Quarterly Reports on Form 10-Q and other filings with the Securities and Exchange Commission. Copies of these filings are available online at www.sec.gov, www.jnj.com, www.investor.jnj.com or on request from Johnson & Johnson. Johnson & Johnson does not undertake to update any forward-looking statement as a result of new information or future events or developments.
1 Dr. Sharma served as a study investigator and as a consultant for Johnson & Johnson. Dr. Sharma was not compensated for this authorship contribution. 2 Dr. Natale served as a study investigator and as a consultant for Johnson & Johnson. Dr. Natale was not compensated for this authorship contribution. 3 Dr. Mansour served as a study investigator and as a consultant for Johnson & Johnson. Dr. Mansour was not compensated for this authorship contribution. 4 Dr. Porterfield served as a study investigator and as a consultant for Johnson & Johnson. Dr. Porterfield was not compensated for this authorship contribution. 5 Dr. Jazayeri served as a study investigator and as a consultant for Johnson & Johnson. Dr. Jazayeri was not compensated for this authorship contribution. 6 Dr. Zou served as a study investigator and as a consultant for Johnson & Johnson. Dr. Zou was not compensated for this authorship contribution. 7 J&J MedTech US EP Market Dynamics. Source: DRG Clarivate. Data Latency: 8 weeks. Market Coverage: ~35% US Hospitals.
Footnotes i Weisman D, Khanna R, Maccioni S, Rong Y, Al-Azizi KM. Pulsed field ablation using a large-tip focal catheter with 3D mapping integration: early outcomes from the OMNY-AF single-arm pilot study. Presented at: AFib Symposium; February 6, 2026; Boston, MA. ii Mansour M, Michaud G, Di Biase L, Zei P, Sauer W, Heist K, Nair D, Reddy V, Natale A. Reduced neurovascular events following workflow and irrigation adjustments with a variable loop circular catheter for pulse field ablation. Presented at: AFib Symposium; February 5–7, 2026; Boston, MA. iii Porterfield C, Munjal J, Hushion M, Varley A, Haas P, Quin EM, Rouse C, Krishnan K, Marrouche N. The variable loop circular catheter safety survey (VARISURE): early results. Presented at: AFib Symposium; February 5-7, 2026; Boston, MA. iv Jazayeri M, Khaykin Y, Morales G, Joshi N, Silva J, Hughey A, Steckman D, Osorio J, Zei P, Koplan B, Silverstein J, Ebinger M, Greenberg J, Dominic P, Conti S, Quadros K, Saleem M, Smith M, Gampa A, Porterfield C, Krishnan K. Acute safety profile of variable loop circular catheter pulsed field ablation for paroxysmal and persistent atrial fibrillation in the REAL AF registry. Presented at: AFib Symposium; February 5-7, 2026; Boston, MA. v Zou F, Zhang X, Gomez T, Byun E, Chen Q, Marazzato J, Schiavone M, Mohanty S, La Fazia VM, Motta J, Zamora C, Pandey S, Safren L, Safren Y, Grupposo V, Ynoa D, Lin A, Natale A, Guttenplan N, Di Biase L.Irrigation flow optimization during pulsed field ablation: preclinical insights with a variable loop circular catheter (VLCC). Presented at: AFib Symposium; February 5-7, 2026; Boston, MA. viA Study of Assessment on Safety and Effectiveness of BWI Pulsed Field Ablation With OMNYPULSE Catheter for the Treatment of Paroxysmal Atrial Fibrillation (PAF) (OMNY-AF). Clinicaltrials.gov. Accessed January 30, 2026. vii Jnjmedtech. OMNYPULSETM Bi-Directional Catheter IFU.
Cloud-based alerting systems often struggle to distinguish between normal cloud activity and targeted malicious operations by known threat actors. The difficulty doesn’t lie in an inability to identify complex alerting operations across thousands of cloud resources or in a failure to follow identity resources, the problem lies in the accurate detection of known persistent threat actor group techniques specifically within cloud environments.
In this research, we hypothesize how a new method of alert analysis could be used to improve detection. Specifically, we look at cloud-based alerting events and their mapping to the MITRE ATT&CK® tactics and techniques they represent. We believe that we can show a correlation between threat actors and the types of techniques they use, which will trigger specific types of alerting events within victim environments. This distinct, detectable pattern could be used to identify when a known threat actor group compromises an organization.
To prove this method of alert analysis, Unit 42 researchers focused on two known threat actor groups that use two fundamentally different types of operational techniques to compromise their victims’ cloud environments. These groups are the cybercrime group Muddled Libra and the nation-state group Silk Typhoon. Both threat actor groups are known to target cloud operations.
We analyzed cloud alerting events across 22 industries between June 2024 and June 2025.
The research was conducted by pairing the cloud-related MITRE ATT&CK techniques known to be used by Muddled Libra and Silk Typhoon with the specific security alerts they are known to trigger in cloud environments.
The test confirmed, as you will see within the remainder of this article, that security teams can successfully distinguish unique alerting patterns between Muddled Libra and Silk Typhoon based solely on the types of alerts observed.
Additionally, the results show a clear link between threat actors’ cloud-focused operations and the industries those groups target. Therefore, at times when one of the groups was known to be attacking certain industries, we can see those patterns appear in our data.
The confirmation that our detection method works as expected opens the door to the possibility of automated prevention capabilities for complex cloud architectures.
Cortex Cloud is designed to detect and prevent the malicious operations, configuration alterations and exploitations discussed in this article, by associating events with MITRE tactics and techniques. These capabilities help organizations to maintain runtime detection of events.
Organizations can gain help assessing cloud security posture through the Unit 42 Cloud Security Assessment.
If you think you might have been compromised or have an urgent matter, contact the Unit 42 Incident Response team.
Another Lens on Cloud Alert Trends
Following our previous article on cloud alert trends, we conducted another analysis of cloud alert statistics.
As part of the effort to determine whether we could identify threat groups, this time we analyzed the data in terms of the industries in which cloud alerts were triggered. Adding industry telemetry to the analysis allowed us to focus our efforts on identifying the techniques, and thus the resulting alerts, used by these threat actors as a control parameter. Using alert data pulled between June 2024 and June 2025, we identified the industries that saw the highest number of unique alert types as well as the highest average number of daily alerts. We then correlated these trends with the activities and targets of two threat groups: Muddled Libra and Silk Typhoon.
This article presents our analysis of Muddled Libra and Silk Typhoon operational techniques and the associated alert analysis.
Glossary: Mapping Techniques to Alerts
The research was conducted by analyzing cloud-related MITRE ATT&CK techniques known to be used by Muddled Libra and Silk Typhoon and pairing them with the specific security alerts they are known to trigger in cloud environments. The following glossary will assist readers in understanding the results we present.
Mapping MITRE Techniques to Alerts: A single MITRE technique can potentially trigger multiple unique security alerts, and conversely, a single alert can map to one or more MITRE techniques and tactics. For example, the alert Remote command line usage of serverless function’s token in the Cortex Cloud platform correlates to the MITRE tactic Credential Access, and the MITRE techniques Steal Application Access Token and Unsecured Credentials.
Unique Alert Count: We counted each alert rule only once for the basis of this research. For example, we identified nearly 70 different unique alerting rules that could be attributed to at least one of the 11 different cloud-related MITRE techniques known to be used by Muddled Libra. For Silk Typhoon, we found just over 50 unique alerting rules that could be attributed to at least one of their 12 known cloud-related MITRE techniques. Additionally, we found that only three unique alert rules were present in both Muddled Libra and Silk Typhoon alert rule sets. In some cases, these alerting rules triggered multiple times within our data across multiple organizations, but when we refer to unique alerts within an industry, we are only considering whether an alert triggered at all during the specified period.
Average Daily Occurrences: If a threat actor used the MITRE technique Data from Cloud Storage (T1530), one of the resulting unique Cortex alert rules might be Suspicious identity downloaded multiple objects from a bucket. If this alert is triggered 1,000 times in a single day, it counts as a single unique alert, but the 1,000 occurrences of that alert in that day will be calculated in the average daily occurrences. When we report average alerts per day by industry in the article below, we take the average for each organization within that industry vertical.
To use a metaphor to help explain how we considered alerts, if each alert rule was a type of fruit, we would see that Muddled Libra holds a very different basket of fruit than Silk Typhoon does. In fact, the baskets are so diverse, that out of the nearly 70 different types of fruit Muddled Libra has, and the more than 50 different types of fruit Silk Typhoon has, they only have 3 types of fruit in common.
When we look at alerts triggered within an industry, we might see a variety of fruit scattered about — maybe 10 oranges, 14 lemons and so on. When we analyze the fruit trail in terms of the types of fruit found within a particular industry, compared with the types of fruit found in the baskets we know Muddled Libra or Silk Typhoon to be holding, we can make a reasonable determination of which threat actor was involved.
Methodology
We collected alerts between June 2024 and June 2025 that were triggered on a combination of platforms, including:
Cloud service providers
Container environments
Cloud-hosted applications
SaaS platforms
We then analyzed the alerts based on their unique naming, originating platform, alert date and metadata such as:
Industry
Region
Frequency of occurrence
Average number of occurrences in each organization
As described above, we integrated the correlation of the MITRE ATT&CK framework, by pairing each alert with its corresponding MITRE technique.
We also analyzed the correlation between the targeted organization’s industry and region and the severity level of the alerts they experienced. This helped to identify the types of alerts that are more likely to occur, based on these factors.
Threat Actor Profiles
Muddled Libra
Background
Muddled Libra (also known as Scattered Spider, or UNC3944) is a cybercrime group that has been active since 2021.
Known for its use of social engineering, including making calls to organizations’ help desks, Muddled Libra has also been known to partner with ransomware-as-a-service (RaaS) programs. By continually updating its approach, the group has successfully used social engineering techniques, including smishing (SMS phishing), vishing (voice phishing) and spear phishing (directly targeting an employee).
Upon successfully compromising an organization, the group uses several tools, including ransomware variants such as DragonForce – a subscription-based RaaS framework created by a group of the same name, tracked by Unit 42 as Slippery Scorpius. The group also uses cloud enumeration tools such as ADRecon, an open source Active Directory reconnaissance tool.
Targeted Industries and Techniques
While Muddled Libra’s targeted industries have evolved since 2022, the following sectors have been consistently reported:
Aerospace and defense
Financial services
High technology
Hospitality
Media and entertainment
Professional and legal services
Telecommunications
Transportation and logistics
Wholesale and retail
Muddled Libra employs multiple offensive techniques to compromise and maintain access within a victim’s environment. We analyzed the group’s known techniques, and extracted those techniques that specifically focus on cloud infrastructure, as Table 1 shows. Together, these form a sort of “fingerprint” that we can use to identify the group within cloud alert data.
Table 1. Known Muddled Libra cloud tactics and techniques.
Methodology Walkthrough
Even though each MITRE Technique is relatively granular in terms of scope of operation, there can be multiple types of computational events from a cloud platform or software-as-a-service (SaaS) application which can fall under the purview or scope of a single MITRE technique.
For example, the MITRE technique T1078.004 – Valid Accounts: Cloud Accounts is focused on the operational event of a valid cloud account. This can have a wide scope in the types of event which can be counted, such as:
Unusual resource modification from a newly seen IAM user
Deletion of multiple cloud resources by a newly created IAM role
A suspicious identity created or updated password for an IAM user
Each of these can be linked to a valid cloud account but each one could have vastly different root causes.
Additionally, when looking specifically at an individual alert type, such as Unusual resource modification from a newly created IAM role, this event could be considered to align not only with the MITRE tactic Initial Access, but it could also align with the MITRE tactics Defense Evasion or even Persistence.
When we expanded our scope to include potential alerting events that could be triggered by any of the MITRE techniques known to be used by Muddled Libra, we found nearly 70 alerting events that could be attributed to at least one of these MITRE techniques.
We collected all of these alerts, which were associated with each of the MITRE techniques known to be used by Muddled Libra. We then distilled those alerts to identify the number of unique alerts that were present within each industry. We also tracked the number of average daily occurrences for each organization within those industries. To use our fruit analogy, we identified the number of unique fruit that the threat actors left at each respective industry (unique alerts), then we also counted how many of each fruit type were present at each organization within that industry (average alert count).
As explained in the Glossary section, we were able to use these numbers to build patterns.
Industry and Technique Analysis
Comparing the triggered alerts and their associated MITRE techniques with the targeted industries shows a correlation between the industries targeted according to public reports and the alerts triggered by Muddled Libra operations. Figure 1 shows this correlation by ranking industries from the highest to the lowest based on the number of unique alerts related to the MITRE techniques listed within Table 1, between June 2024 and July 2025. Industries that were publicly reported as targeted are shown in red.
Figure 1. Count of unique alerts by industry from June 2024-June 2025. Red bars indicate the industries publicly reported as targeted by Muddled Libra.
Figure 2 shows the average daily number of alerts that occurred during the same timeframe.
Figure 2. Count of the average daily alerts by industry from June 2024-June 2025. Red bars indicate the industries publicly reported as targeted by Muddled Libra.
Figure 2. Count of the average daily alerts by industry from June 2024-June 2025. Red bars indicate the industries publicly reported as targeted by Muddled Libra.
While the highest volume of unique alerts (shown in Figure 1) aligns perfectly with Muddled Libra’s most-reported targets — specifically high technology, wholesale and retail, financial, and professional and legal services—the presence of a number of signature alerts in other sectors shouldn’t be ignored. When an industry like manufacturing or pharma and life sciences or state and local government shows a significant subset of Muddled Libra’s “fingerprint” (for instance, 16 or more unique alert types), it suggests the group could have an active interest in these environments even if we haven’t seen headlines about it. Security teams in these “middle-tier” industries should treat these clusters of unique alerts as early warning signs that these industries are witnessing a significant number of the group’s known operational techniques.
The unique alert data (Figure 1) should be considered alongside average daily alert data (Figure 2) to distinguish between a threat actor’s strategic breadth and their operational persistence. For instance, transportation and logistics serves as a primary example of high-intensity targeting; it ranks sixth in unique alert variety but first in average daily volume, showing a 25% spike in unique alerts in June 2025 alone. This combination indicates that Muddled Libra is not only using a wide array of its signature techniques in this sector but is doing so with higher frequency. We will take a deeper dive into transportation and logistics in the next section.
In contrast, telecommunications and media and entertainment were some of the first and most frequent targets of Muddled Libra in 2022 and 2023, but their standings as the last two positions for average daily alerts in 2024-2025 suggest that these two industries in particular have experienced a saturation effect. Namely, the targeting of these groups may be aging off. They no longer appear to be the key focus of the Muddled Libra actors. The other industries that could also fall into this category are hospitality and aerospace and defense.
For a defender, this data provides a threshold for proactive investigation. A high count of unique alerts (the “variety” of the fruit basket) typically signals a sophisticated, multi-stage intrusion attempt, whereas a high daily average (the “quantity” of fruit) may point to automated scanning or persistent credential stuffing. If your organization sees more than 10 unique Muddled Libra-associated alerts within a 30-day window, it is time to look deeper, regardless of whether your specific industry is currently “trending” in threat intelligence circles. The goal is to move from reactive patching to proactive defense by identifying these actor-specific patterns before they escalate to data exfiltration.
Focused Analysis: Aviation
Reports that Muddled Libra was targeting the aviation industry initially surfaced in June 2025. Unit 42 does not track aviation as a singular category. Instead, aviation organizations appear under our transportation and logistics category.
When looking at the transportation and logistics industry alerts, we found an increase in the number of unique alerts based on the MITRE techniques used by Muddled Libra during this same timeframe.
It is important to note that here we are looking at the number of unique alert rules for this analysis and breaking this out by month, as opposed to the whole year view shown in Figure 1. We arrive at “unique alerts” for the industry by taking the average number of unique alerts seen for each organization tracked in that category over a monthly timeframe.
Figure 3 shows that the average number of unique alerts per organization in the transportation industry increased by 25% from May-June 2025. What makes this finding important is that Muddled Libra made several headlines during June 2025 for its operations targeting airline organizations.
Figure 3. Unique alerts by month for the transportation industry. The bar for June is red because it is the period publicly reported to have the highest targeting of the aviation industry.
As illustrated in Figure 3, June 2025 saw the highest number of unique alerts for the transportation industry, with 15 unique alerts.
The Verdict on the Fingerprinting Effort
Looking at the correlation, there does appear to be a fingerprint capability that could be used as a detection pattern. This pattern could help organizations identify if they are potentially targeted and take mitigative steps. Additionally, this could also assist organizations in developing an early warning detection trigger. For example, if defenders witness an increase in the number of daily average occurrences alerts from known Muddled Libra techniques, this could indicate reconnaissance or discovery activity occurring against their infrastructure. This then provides an opportunity to proactively prepare for future operations.
Top 10 Alerts from Muddled Libra Techniques
Table 2 lists the top 10 alerts that we observed in association with Muddled Libra’s MITRE techniques.
Alert Names
MITRE Techniques
Tactics
Azure sensitive resources enumeration activity using Microsoft Graph API
T1526
Discovery
Microsoft 365 storage services exfiltration activity
T1530
Collection, Exfiltration
Multi region enumeration activity
T1580
T1535
T1526
Discovery
Storage enumeration activity
T1619
T1530
T1526
Discovery, Collection
Cloud Identity Queried Cost or Usage Information
T1087.004
T1580
Discovery
A cloud identity invoked IAM related persistence operations
Table 2. The top 10 alerts associated with Muddled Libra’s MITRE techniques.
As outlined in Table 2, Muddled Libra has an extensive history of targeting Microsoft Azure environments using Graph API, a RESTful API that enables access to Azure cloud resources. This type of activity correlates with the MITRE techniques used by Muddled Libra and the alerts triggered by their operations. The most frequent alert between June 2024 and June 2025, in relation to the MITRE techniques used by Muddled Libra, was resource enumeration using Microsoft Graph API. The next most common alert was for exfiltration activity from Microsoft 365 storage services. While discovery operations represented the bulk of the remaining alert types, collection and exfiltration operations were the second most frequent alert type.
Silk Typhoon
Background
Silk Typhoon (also known as HAFNIUM) is a China-nexus threat actor group that has been in operation since at least 2021. This group has historically exploited multiple vulnerabilities on Microsoft Exchange Servers. In recent years, the group appears to be shifting targets towards cloud environments, using compromised credentials obtained via vulnerable public-facing VPN endpoints to move laterally through cloud environments. The group relies on remote monitoring and management (RMM) tools to maintain persistent access and leverages Microsoft’s Graph API to enumerate cloud resources.
Targeted Industries and Techniques
Cybersecurity researchers have identified that the industries most commonly targeted, primarily located within the U.S., include:
Education
High technology
Federal governments
Financial services
Nongovernmental organizations (NGOs)
Professional and legal services
State and local governments
Utilities and energy
Silk Typhoon has employed several offensive techniques to compromise and maintain access within a victim’s environment. Using the same methodology that was employed during the Muddled Libra analysis above, we analyzed the techniques and identified those that focus on cloud infrastructure, as Table 3 shows. We found that between Silk Typhoon and Muddled Libra’s known employed technique usage, only three techniques were employed by both threat actor groups, T1530, T1078.004 and T1098.001. This provides a basis on which to compare and contrast the results between both groups’ operations and, more importantly, on the types of alerts witnessed by organizations in the industries they target.
MITRE Tactics
MITRE Techniques
MITRE Technique Name
Collection
T1119
Automated Collection
Collection
T1530
Data from Cloud Storage
Credential Access
T1555.006
Credentials from Password Stores: Cloud Secrets Management Stores
Defense Evasion,
Lateral Movement
T1550.001
Use Alternate Auth Material: Application Access Token
Defense Evasion,
Persistence,
Privilege Escalation,
Initial Access
T1078.004
Valid Accounts: Cloud Accounts
Discovery
T1619
Cloud Storage Object Discovery
Exfiltration
T1567.002
Exfiltration Over Web Service: Exfiltration to Cloud Storage
Table 3. Known Silk Typhoon cloud tactics and techniques.
Methodology Walkthrough
As a brief recap to the methodology of our research, we analyzed the types of alerting events that could be associated with each of the MITRE Techniques known to be used by Silk Typhoon. When we included potential alerting events that could be triggered by any of the MITRE Techniques known to be used by Silk Typhoon, we found just over 50 alerting events that could be attributed to at least one of these MITRE Techniques.
We collected all of these alerts and distilled those alerts to identify the number of unique alerts that were present within each industry and the number of average daily occurrences of those alerts for each organization within those industries.
To use the same analogy as above, we wanted to identify what types of fruit Silk Typhoon brought to the party, and how many pieces of fruit they typically deploy when attacking.
Industry and Technique Analysis
We compared the total number of unique alerts for each month from June 2024 to June 2025 with the industries from which those alerts were triggered. This comparison confirmed that we were able to see the “fingerprints” of Silk Typhoon in the alerts triggered in industries that the group was known to be targeting.
As mentioned, Silk Typhoon had just over 50 unique alerts associated with their known technique usage, where Muddled Libra had nearly 70.
In contrast, we saw higher numbers of unique alerts within each industry when examining our Silk Typhoon data than we did in our Muddled Libra data.
In other words, Silk Typhoon may be holding a basket with fewer types of fruit (50) than in Muddled Libra’s (70), but the threat actor seems to use more types from the basket in its operations (i.e. as many as 27 unique alerts as opposed to 22).
The graph in Figure 4 shows our observations of alerts by industry over the period studied.
Figure 4. Count of unique alerts and average daily alerts by industry. Red bars indicate the industries public reported as targeted by Silk Typhoon.Figure 5. Count of the average daily alerts by industry. Red bars indicate the industries publicly reported as targeted by Silk Typhoon.
While the highest volume of unique alerts aligns with Silk Typhoon’s most-reported targets—specifically high technology, financial services, and professional and legal services—the presence of signature alerts in other sectors is equally telling. When an industry like wholesale and retail or manufacturing shows a significant subset of Silk Typhoon’s “fingerprint” (for instance, 18 or more unique alert types), it indicates that the group could be actively deploying their offensive techniques against these industry environments. This could be occurring even if public reporting is minimal or nonexistent at this date. Security teams in these “middle-tier” industries should treat these clusters of unique alerts as evidence that they are witnessing a broad spectrum of the group’s known operational techniques, rather than isolated incidents.
The unique alert data for Silk Typhoon (Figure 4) should be considered alongside average daily alert data (Figure 5) to distinguish between a threat actor’s strategic breadth and their operational persistence. To return to our metaphor, Silk Typhoon holds a “basket” with fewer types of fruit (50) than Muddled Libra (70), but they tend to use more of what is in their basket at any given time. For example, we witnessed as many as 27 unique alerts in a single sector compared to Muddled Libra’s 22.
The federal government serves as a primary example for high-intensity targeting. This industry ranks last in the number of unique alerts, or rather in variety, ie. the “types of fruit” in its basket, but first in average daily volume, peaking at 7.28 alerts per day (the “quantity of fruits witnessed”). This suggests that while Silk Typhoon may use a narrower set of techniques against government targets, it deploys those specific tactics with relentless frequency. Conversely, high technology shows a “worst-of-both-worlds” scenario, ranking first in unique tactical variety and near the top for daily volume. This indicates campaigns that are both sophisticated and persistent.
Similar to our comments on unique alerts, when we see a high level of activity possibly related to the threat group, it may be worth defenders’ threat hunting for other known alerts related to Silk Typhoon, out of an abundance of caution. High levels of average alert activity could signify threat groups trying to gain initial access, but not yet succeeding in deploying their full toolset.
For a defender, this data provides a threshold for proactive investigation: a high count of unique alerts (the “variety” of the fruit basket) typically signals a sophisticated, multi-stage intrusion attempt, whereas a high daily average (the “quantity” of fruit) may point to automated scanning or persistent exploitation of specific vulnerabilities. If an organization observes more than 10 unique Silk Typhoon-associated alerts within a month, it is time to look deeper, regardless of whether a specific sector is making headlines as a common target.
Top 10 Alerts from Silk Typhoon Techniques
Table 4 lists the alerts most commonly seen in relation to the MITRE techniques used by Silk Typhoon.
Alert Names
MITRE Techniques
Tactics
Microsoft O365 storage services exfiltration activity
T1530
Collection, Exfiltration
Process execution with a suspicious command line indicative of the Spring4Shell exploit
T1190
Initial Access
Storage enumeration activity
T1619
Discovery
A cloud identity invoked IAM related persistence operations
T1098
Persistence, Privilege Escalation
Suspicious identity downloaded multiple objects from a bucket
T1530
T1020
Collection, Exfiltration
Suspicious identity downloaded multiple objects from a backup storage bucket
T1530
T1020
Collection, Exfiltration
An identity performed a suspicious download of multiple cloud storage objects
T1530
T1020
Collection, Exfiltration
An identity performed a suspicious download of multiple cloud storage objects from multiple buckets
T1530
T1020
Collection, Exfiltration
Massive code file downloads from SaaS service
T1530
Collection
Deletion of multiple cloud resources
T1485
Impact
Table 4. The top 10 alerts associated with Silk Typhoon’s MITRE techniques.
As outlined in Table 4, collection and exfiltration techniques were the most common alerts associated with Silk Typhoon’s MITRE techniques. Microsoft 365 storage services exfiltration was the most frequently observed alert. Other alerts identified include cloud storage enumerations and suspicious downloads of cloud storage objects.
Industry Cloud Alert Trends
Perhaps the most striking result of our research came to light when we compared general cloud alerting trends with the trends we discovered while performing the fingerprinting analysis of Muddled Libra and Silk Typhoon.
The industry of high technology was consistently the top ranked industry when considering general cloud alerting trends, as well as the two threat actor groups’ alerting trends. However, the remaining industries we studied did not follow a uniform pattern. As shown in Figure 6, we found that the order of the most targeted industries shifted when we only counted alerts from Muddled Libra and Silk Typhoon operations.
Figure 6. Ranking of the top industries by unique alert counts for all alerts, Muddled Libra and Silk Typhoon.
As stated above, the high technology industry is in first place across both threat actor group findings, as well as the top ranking industry when looking at all cloud alerts by industry.
However, the remaining industry ranks do not mirror the same results. Looking at the wholesale retail industry illustrates a key finding. This industry is the second highest for Muddled Libra alerting events and third highest for Silk Typhoon, but is 14th on the industry list for all alerts.
This indicates that the fingerprinting analysis on these alerting operations does not reflect the same pattern as the general noise of all alerting trends. It appears that the distinct operations performed by the threat actors against the industries they target carry their own unique trends.
Conclusion
Our analysis confirms the capacity to leverage the alerts triggered as a fingerprint detection pattern for the malicious techniques used by Muddled Libra and Silk Typhoon. This distinct detection capacity offers a new pathway for organizations to implement predictive and proactive cloud defense strategies.
Our research successfully differentiated the MITRE tactic and technique operations used by Muddled Libra, notably the aviation industry’s 25% increase in the number of unique alerts compared to the previous month, and Silk Typhoon’s increased higher than average number of daily alerts within the Federal and State Government industry.
By identifying the alert patterns that each threat actor’s techniques have on the alerting events within cloud environments, threat researchers can identify the threat actors most likely to target certain industries using specific techniques. This can then help defenders to proactively prepare defenses against those types of threats. Through the analysis of threats based on the types of attack techniques they leverage, organizations can create a defense methodology built specifically for their industry vertical.
Proper implementation of these defense controls can be effective in defending against targeted threat actor scenarios through the creation of tailored defensive alerting. Additionally, these controls can provide the capability to detect early warning scenarios and techniques, such as initial access operations, enabling prevention operations to block malicious cloud operations before they escalate to execution, impact or exfiltration.
Palo Alto Networks customers are better protected from the threats discussed above through the following products:
Cortex Cloud customers can help secure and protect their cloud environments through compliance guardrails, application security monitoring and prevention techniques and through the proper placement of Cortex Cloud XDR endpoint agent and serverless agents within a cloud environment. Cortex Cloud is designed to identify cloud events witnessed on cloud platforms, to protect cloud posture and runtime operations. By associating events with MITRE tactics and techniques, Cortex Cloud helps detect and prevent the malicious operations, configuration alterations and exploitations discussed within this article.
Organizations can gain help assessing cloud security posture through the Unit 42 Cloud Security Assessment.
If you think you may have been compromised or have an urgent matter, get in touch with the Unit 42 Incident Response team or call:
North America: Toll Free: +1 (866) 486-4842 (866.4.UNIT42)
UK: +44.20.3743.3660
Europe and Middle East: +31.20.299.3130
Asia: +65.6983.8730
Japan: +81.50.1790.0200
Australia: +61.2.4062.7950
India: 000 800 050 45107
Palo Alto Networks has shared these findings with our fellow Cyber Threat Alliance (CTA) members. CTA members use this intelligence to rapidly deploy protections to their customers and to systematically disrupt malicious cyber actors. Learn more about the Cyber Threat Alliance.
You don’t have permission to access “http://www.imf.org/en/publications/technical-assistance-reports/issues/2026/02/06/tonga-technical-assistance-report-liquidity-management-and-forecasting-573683” on this server.
A trader works as the Dow Jones Industrial Average surpasses the 50,000 mark on the floor at the New York Stock Exchange (NYSE) in New York City, U.S., Feb. 6, 2026.
Brendan McDermid | Reuters
Stocks surged on Friday as technology stocks recovered following several days of heavy selling in the sector and bitcoin rebounded following a rout that took the popular cryptocurrency down more than 50% at one point.
The Dow Jones Industrial Average advanced 1,206.95 points, or 2.47%, closing at 50,115.67. Friday marked the first time the Dow exceeded the 50,000 level. The S&P 500 jumped 1.97% and ended at 6,932.30, while the Nasdaq Composite advanced 2.18% to 23,031.21. With those moves, the S&P 500 climbed back into the green for 2026.
Even with Friday’s pop, the S&P 500 posted a 0.1% decline for the week, while the Nasdaq fell 1.8% on the week. The 30-stock Dow rose 2.5% week to date, benefiting from some rotation into some economically cyclical stocks even as the overall market was weighed down by tech selling.
Dow Jones Industrial average, 5 days
Nvidia and Broadcom were two of the key winners Friday, with the former increasing by nearly 8% and the latter growing 7% following big declines earlier in the week. Other stocks such as Oracle and Palantir Technologies also bounced back as investors reconsidered some of the names at cheaper levels. Oracle and Palantir each rose 4%. Some key software stocks like ServiceNow — which has been the epicenter of the tech sell-off because of an artificial intelligence disruption fear of software — remained weak on Friday, however.
“We’re in a gold rush right now with AI,” said Falcon Wealth Planning founder Gabriel Shahin.
“You have the investment that Google is making, Nvidia is making, that Meta is making, that Amazon is making. There is money that will be deployed,” he also said. “It’s just the carousel [of money movement] sometimes scares people.”
Shahin believes the market is in the midst of a “great recalibration,” where investors are going to move further out of growth stocks and into value. Over the coming months, his bet is on large-cap value names. That played out Friday, with investors buying up shares in areas such as industrials and financials. In those sectors, Caterpillar and Goldman Sachs were standouts, supporting the Dow’s outperformance with their rise of 7% and 4%, respectively. Small-cap stocks also saw a boost, with the Russell 2000 index rallying 3.6%.
Bitcoin recouped some losses Friday, adding 10% and touching a session high of $71,458.01 after briefly sinking below $61,000 overnight to its lowest level since October 2024 — more than 52% off from its record high of $126,000 hit in early October 2025. Friday’s move higher helped ease some of the risk-off concerns among investors that recently plagued the broader market. The cryptocurrency has lost 16% this week, however.
The week was bleak heading into Friday, with the S&P 500 on pace for its worst week since last October and the Nasdaq Composite on track for its worst week since the tariff-related market plunge of last April. Friday’s pop pared those declines significantly.
Amazon was an outlier Friday, as shares sank more than 5% after the e-commerce giant posted earnings per share slightly under analyst expectations and told investors to expect $200 billion in capital expenditures this year.
Albert Bourla, chairman and CEO of Pfizer, speaks at The Wall Street Journal’s Future of Everything Festival in New York City, U.S., May 22, 2024.
Andrew Kelly | Reuters
A version of this article first appeared in CNBC’s Healthy Returns newsletter, which brings the latest health-care news straight to your inbox. Subscribe here to receive future editions.
Pfizer made one thing clear this week: It’s officially back in the obesity race.
The drugmaker is laser-focused on bringing to market treatments from its $10 billion acquisition of the obesity biotech Metsera. On Tuesday, it released promising phase two trial data on one injection, called PF′3944, that’s furthest along in development.
The experimental drug drove solid weight loss when taken once a month in a mid-stage trial – offering early evidence that the injection can be administered less frequently than existing drugs without sacrificing efficacy. That could be a boon to Pfizer after it faced several setbacks in trying to win a slice of a market dominated by weekly injections from Eli Lilly and Novo Nordisk, along with Novo’s new daily pill.
Patients with obesity or who are overweight lost up to 12.3% of their weight compared with placebo at week 28 in the ongoing phase two study. The company said no plateau was observed after patients transitioned to monthly dosing, which suggests that continued weight loss is expected as the study continues through week 64.
But investors are still looking for the full data from that trial, which is slated to be presented at a medical conference in June. Pfizer also plans to start 10 phase three studies on the injection, with the goal of achieving the first of several potential approvals in 2028.
I talked to Pfizer CEO Albert Bourla and other top execs about the data this week and the company’s broader obesity strategy. Here’s what they had to say.
A potential “best-in-class” product
Bourla told CNBC that the data shows the monthly product has a “very competitive profile in tolerability and efficacy.”
Pfizer plans to use a higher dose of the drug in phase three trials, and Bourla said it will produce efficacy and tolerability data that is “maybe best in class, so better than anything else,” while being taken less frequently. The company’s modeling predicts that the higher dose could result in 16% weight loss at week 28.
In the phase two trial, patients started on weekly injections of the drug for 12 weeks before switching to once-monthly dosing.
Pfizer also plans to study people who are taking existing weekly GLP-1s and give them the option to switch to the company’s monthly shot, said Dr. Jim List, Pfizer’s chief internal medicine officer.
List said that’s one of the selling points of the company’s injection: it can serve as a more convenient maintenance treatment for patients to switch to.
“If you say, listen, I can give you one of these drugs. This one, you’ll take once a week for the rest of your life. But this other one, you’ll take once a week, and you could also switch it to once a month. Which one do you want?” List said. “It’s always going to be the one with more options.
He added that “weekly doesn’t work for everybody,” since some patients need to travel and can’t keep their injections refrigerated.
Bourla said people who have been taking weekly injections are also more likely to switch to another shot rather than an oral option.
“The oral will be for people, but they didn’t start with the needle,” he said. “So as a result, I think the monthly or longer products will probably become a standard, and we are the first and hopefully the best.”
Combination regimens
A key part of Pfizer’s strategy for the PF′3944 injection is to combine it with another drug targeting a gut hormone called amylin, List said.
“We’re expecting to get even more weight loss with that combination than we get with this GLP-1 alone,” he said.
Amylin is hormone co-secreted with insulin in the pancreas to suppress appetite and reduce food intake. Amylin treatments have a similar effect to GLP-1s like Lilly’s Zepbound and Mounjaro, but some analysts and researchers say it could be easier for patients to tolerate and help them preserve lean muscle mass.
Pfizer on Tuesday said early data showed that the two drugs together caused an additive weight loss of 5% when compared to placebo at day 8. Amylin alone also showed weight loss of 8.4% at day 36.
Both drugs are ultra-long-acting, meaning they are engineered to remain active in the body for longer than existing treatments like Novo’s Wegovy and can be taken once a month.
Pfizer plans to share more data on the amylin drug during the medical conference in June. List said the company is advancing the product into phase two trials in the first half of this year.
Quarterly dose GLP-1 injection
Pfizer on Tuesday also teased a potential GLP-1 injection that is dosed quarterly – once every three months – rather than monthly or weekly.
List said the injection will be “ultra-ultra-long-acting,” so Pfizer will be finding a way for the drug to have “slower degradation in the human body so that it can certainly last longer” than PF′3944.
Chief Scientific Officer Chris Boshoff told CNBC that the vast majority of patients will prefer an injection, and “obviously, being monthly will be preferable over weekly, and likely three-monthly maybe better than monthly.”
But List said it’s still early days for that drug.
Feel free to send any tips, suggestions, story ideas and data to Annika at a new email: annika.constantino@versantmedia.com.
Earn‑outs are a familiar tool in M&A transactions, often helping bridge valuation gaps by tying part of the purchase price to the future performance of the business. But they also generate some of the most common post‑closing disputes, especially around whether certain buyer actions trigger early payment of the remaining earn‑out.
The Ontario Superior Court of Justice’s 2025 decision1 and the Court of Appeal’s 2025 confirmation in Project Freeway Inc. v. ABC Technologies Inc.2 provide practical guidance on how courts may interpret earn‑out acceleration clauses and the extent to which pre‑closing documents, such as letters of intent (LOIs), can influence that interpretation. See our summary of the case and its key takeaways below.
The dispute
Project Freeway Inc. sold its business to ABC Technologies Inc. under a share purchase agreement (SPA) that included a potential US$26.4 million earn‑out and an acceleration clause requiring immediate payment of any remaining earn‑out if ABC sold a “material portion” of the business’s assets to a non‑affiliate without the seller’s consent. Project Freeway and ABC had, prior to entering into the SPA, entered into a non-binding LOI in respect of the transaction.
After closing, ABC completed two transactions without Project Freeway’s consent: (a) a sale‑leaseback of major operating real estate, and (b) an accounts receivable factoring arrangement. Project Freeway asserted that each transaction triggered automatic acceleration. The trial judge disagreed with Project Freeway, and the Court of Appeal affirmed the trial judge’s decision.
What the court decided
The key issue was the interpretation of “a material portion” of the assets. The courts found the phrase ambiguous and applied a contextual and purpose-based approach, focusing on the economic function of the earn‑out rather than a formal, size‑only trigger.
Because the earn‑out was calculated using contribution‑margin metrics, the courts examined whether the post‑closing transactions impaired the business’s ability to meet those targets. They concluded the transactions did not harm the earn‑out regime, and therefore, acceleration was not engaged in the absence of actual economic prejudice to the seller.
Why the LOI still mattered
Although the SPA contained an entire agreement clause, the court considered the LOI and other surrounding circumstances in interpreting the ambiguous term “material,” serving as a reminder that early deal documents can inform the meaning of later provisions in definitive contracts, particularly where the drafting of definitive contracts is ambiguous.
What this means for M&A transactions
Earn‑outs are grounded in economic purpose, not formal triggers; acceleration should not be expected as a windfall.
Vague terms like “material” invite disputes. Courts may interpret “material” in earn out provisions by reference to economic impact on the earn out, not simply quantitative thresholds. If the parties intend size alone to govern, that intention must be explicit.
Early deal documents influence later interpretation despite standard “entire agreement” clauses. Any intention to deviate from those early documents should be made clear in the SPA.
Contracts, both preliminary and final, should be drafted precisely. If specific events (such as sale‑leasebacks and receivables factoring) are meant to trigger the earn-out, this should be expressly stated.
Final thoughts
The Project Freeway decisions underscore that earn‑outs function best when the parties share a clear understanding of their economic purpose and draft the mechanics with precision. In practice, this calls for alignment between preliminary documents (such as letters of intent or term sheets) and the definitive agreement, and where the parties intend to deviate from those preliminary documents, an express indication of that departure in the definitive agreement.
Thoughtful drafting at each stage of the negotiation process remains the most effective way to avoid disputes and preserve the intended economic balance of the deal. The NRF team is available to assist you in this regard.
On January 29, 2026, SEC Chairman Paul Atkins and Commodity Futures Trading Commission (CFTC) Chairman Michael Selig held a rare joint “Project Crypto” summit. Project Crypto is the agencies’ joint effort to modernize the regulatory framework for cryptoassets.
Near the end of his remarks, CFTC Chairman Selig announced his plans to “support the responsible development of event contract markets,” laying out a four-part regulatory agenda.1 Chairman Selig’s remarks signal a new regulatory environment in the short term and a hotly contested regulatory and legal environment in the long term, with important implications for prediction markets, regulators, and market participants.
The key takeaways below distill the most significant themes emerging from Chairman Selig’s remarks and his evolving agenda.
Key Takeaways
The CFTC has shifted its view toward permitting political and sports-related event contracts, withdrawing a prior proposed rule that would have prohibited such contracts.
At the same time, the CFTC is preparing new rulemaking on event contracts, one that is likely to draw industry participation and judicial scrutiny.
Litigation risk is not diminishing — only changing shape, as federal courts, state gaming regulators, and market participants test the boundaries between derivatives regulation and gambling law.
Until the Supreme Court draws the line between event contracts and betting or wagering, prediction markets will operate in a dual regulatory reality — federally legitimized under the CFTC but resisted at the state level in multiple states. And this unsettled question — which has caused most of the pending litigation — will likely demonstrate the seismic shift in administrative law from Chevron to Loper Bright.
The Slow, Then Fast, Development of Prediction Markets
Event contracts have a long and, for the most part, sleepy history. Dating back at least to the 19th century, event contracts present binary yes/no propositions on the outcome of certain future events, such as the closing price of oil, the amount of rainfall during the growing season, or the last freezing day of the year.
Beginning in the 1980s, academic institutions began offering election prediction markets as an alternative form of polling. The Iowa Electronic Markets, launched in 1988 as a small-scale research exchange at the University of Iowa, exemplified the field’s limited early profile with its low stakes and restricted participation. The University of Victoria (New Zealand), likewise in 2014, launched PredictIt, which also offered election prediction markets in the United States, again under rules that restricted its commercial upside.
In recent years, however, platforms such as Kalshi and Polymarket have surged in popularity and visibility, each now handling billions of dollars in monthly trading volume. Although event contracts span a wide variety of subject matters, two have surged in popularity and controversy: (a) political event contracts and (b) sports event contracts. Political event contracts allow individuals to speculate on the outcome of a political event (e.g., who will win an election). Sports event contracts, on the other hand, allow trading on the outcome of a sporting event (e.g., who will win the Super Bowl).
Event contracts operate within the CFTC’s core derivatives framework. These contracts must be listed and traded on CFTC-registered designated contract markets (DCMs) and cleared through a derivative clearing organization (DCO). As a result, event contracts traded on DCMs are subject to the full panoply of CFTC-market integrity rules and may not be readily susceptible to market manipulation (e.g., wash trading, spoofing, etc.).2
Despite operating within this federally regulated framework, event contracts — particularly those tied to sports outcomes — have become the focus of increasing legal controversy. Much of the litigation on sports-related contracts centers on their perceived similarity to traditional gambling. Indeed, both state and federal courts are currently addressing disputes over whether these sports-related contracts — offered by prediction markets — constitute unlawful sports betting under state gambling laws. That question remains unsettled and may be destined for the Supreme Court.
Against this backdrop of legal uncertainty and active litigation, the CFTC has set a new regulatory agenda for prediction markets.
The CFTC’s New Agenda for Prediction Markets
1. Withdrawal of Prior Actions
Chairman Selig directed CFTC staff to withdraw the 2024 proposed rule that would have prohibited political and sports-related event contracts.3 Days later, the CFTC formally announced the withdrawal in a staff letter, explaining that developments in the event contracts market since the rule’s proposal had rendered it unnecessary.4
In addition, Chairman Selig directed CFTC staff to withdraw a 2025 staff advisory regarding sports-related event contracts. In that advisory, the CFTC cautioned that “state regulatory actions and pending and potential litigation, including enforcement actions, should be accounted for with appropriate contingency planning,” effectively taking a let-the-courts-decide position on whether sports-related event contracts fall within the CFTC’s exclusive jurisdiction.5 By directing staff to withdraw that advisory, however, Chairman Selig signaled a willingness to litigate the position that event contracts — including sports-related ones — belong exclusively within the CFTC’s jurisdiction.
2. New Rulemaking on Event Contracts
During his nomination hearing, Chairman Selig stated that given the complexity of distinguishing between gaming and sports-related event contracts, that determination is best left to the courts — and that the CFTC would follow the law as determined by judicial decisions. But instead of deferring to the courts, Chairman Selig directed CFTC staff to begin drafting a new event contract rule with the goal of establishing clear, workable standards and replacing what he described as a framework that has proven difficult to apply. Chairman Selig’s new agenda comes as the rapid growth of prediction markets shifts from traditional commercial hedgers to retail participants — raising questions about whether a regulatory framework built for institutional commodity markets adequately addresses retail-facing risks.
3. Reassessment of Pending Litigation
Building on its more aggressive jurisdictional posture, the CFTC will reassess its participation in matters currently pending before federal district and circuit courts, particularly where jurisdictional questions are at issue. Jurisdictional issues in pending litigation stem from disputes over whether contracts offered by prediction markets constitute unlawful sports betting under state gambling laws. Although that question remains unsettled,6 Chairman Selig made clear that the CFTC will defend its jurisdiction in ongoing litigation over commodity derivatives.
4. Joint SEC–CFTC Interpretation
Chairman Selig also directed staff to work with the SEC to develop a joint interpretation of Title VII definitions to clarify the boundaries among commodity options, security options, swaps, and security-based swaps. This coordinated effort aims to reduce regulatory uncertainty and avoid innovation’s falling into a “no man’s land” between the two agencies.
Predicting the Legal Landscape of Prediction Markets in Light of the CFTC’s New Agenda
It is impossible to anticipate when these steps will begin or be completed, much less how these developments will affect or be affected by the various prediction market lawsuits that may be destined for resolution by the Supreme Court. However, a few things seem certain.
First, prediction markets that bet on regulatory acceptance of sports-related and political event contracts appear to have been vindicated. Such DCMs as Kalshi and Polymarket may list for trading new contracts by self-certifying with the CFTC that the new contract complies with the Commodity Exchange Act (CEA). Although the CFTC previously declined to take an official position on the permissibility of such self-certifications,7 Chairman Selig’s new agenda suggests that the CFTC is now prepared to do so. That shift comes at a consequential moment: Even before the inauguration, prediction markets had begun offering contracts that increasingly tested the limits of CFTC Rule 40.11’s prohibition on gaming event contracts. And while it remains unsettled whether that regulation prohibits such contracts, it is now settled that the CFTC will not devote its limited resources to taking down sports prediction markets based solely on the subject matter of their contracts.
Second, litigation over prediction markets will continue — and perhaps intensify. Chairman Selig’s statements are likely to embolden both prediction markets to continue investing in the space and state gaming regulators to pursue regulatory action. And the anticipated rulemaking itself may draw legal challenges, regardless of its content.
Third, pending or forthcoming legal challenges will likely demonstrate the seismic shift in administrative law from Chevron to Loper Bright. The CEA’s use of the undefined term “gaming” is sufficiently broad to have granted regulators discretion to adopt broad or narrow definitions, creating the possibility that the CFTC’s interpretation could have vacillated wildly from administration to administration. Under Loper Bright, however, courts will now endeavor to identify the definition of “gaming” rather than the range of reasonable definitions.
Statutory Constraints on the New CFTC’s Agenda: Dodd-Frank as Gatekeeper
Although the CFTC signaled a more permissive approach to prediction markets, its discretion remains bounded by the CEA as amended by the Dodd-Frank Wall Street Reform and Consumer Protection Act. Section 5c(c)(5)(C) of the CEA expressly authorizes the Commission to prohibit event contracts that involve enumerated activities — such as “gaming” — or that are deemed contrary to the public interest.8 So even if the CFTC promulgates a new rule or amends Rule 40.11(a), it will still need to reckon with its statutory mandate under the CEA and comply with the definition of “excluded commodity” under the CEA.9 As a result, any new rulemaking or revision must operate within statutory constraints and will require the CFTC to interpret — not rewrite — the governing provisions of the CEA, a task that will now receive heightened judicial scrutiny in the post–Loper Bright era.
Looking Ahead
Chairman Selig’s remarks mark the most definitive signal yet that prediction markets have moved to the forefront of the CFTC’s regulatory agenda. The rulemaking to follow will draw close attention from market participants, investors, and regulators alike, and any final rule is expected to prompt further litigation. Firms offering, developing, or considering trading event contracts should prepare for increased engagement with the CFTC through comment letters, compliance planning, and strategic regulatory analysis.
Now, as the CFTC looks poised to grant enhanced regulatory legitimacy on sports event contracts and prediction markets more broadly, such a permissive signal has not eliminated state-level resistance, particularly where prediction markets are perceived to encroach on traditional gambling regulation. Days after Chairman Selig’s remarks, and just a few days before the Super Bowl, New York Attorney General Letitia James issued a pointed warning to New Yorkers: “Be careful about placing trades on prediction markets.”10 In doing so, she cautioned that such markets “do not have the same consumer protections as regulated platforms.”11 In her view, prediction markets platforms such as Kalshi and Polymarket “are bets ‘masquerading’ as event contracts,” with the former regulated by state gaming and wagering laws while the latter by the CFTC.12 This warning came on the heels of AG James’s office issuing a cease-and-desist order to Kalshi, alleging that the company’s sports event contracts violated New York’s gambling law.
In parallel, federal prosecutors have issued a warning. The United States Attorney for the Southern District of New York, Jay Clayton, recently stated that his office is actively evaluating how existing fraud statutes apply to prediction markets and expects prosecutions where participants exploit them.13 He emphasized that the prediction-market label does not insulate conduct from criminal liability, including conspiracies to manipulate underlying events.14
Finally, while the CFTC has adopted a new approach to prediction markets, the Commission’s evolving posture should not be understood as a wholesale retreat from enforcement. Even as regulatory focus centers on the permissibility of event contracts themselves, significant questions remain regarding insider trading and the misappropriation of material nonpublic information in prediction markets — especially when the conduct takes place outside the United States, by a non-U.S. citizen, on a non-U.S. platform such as Polymarket. Under CFTC guidance, an entity will generally be considered a “non-U.S. person” if it was formed outside the United States and its officers direct, control, and coordinate the firm’s activities from outside the United States.15 Although classification as a non-U.S. person has important regulatory ramifications for an entity, recent enforcement actions underscore that cross-border structuring alone is not a safe harbor; U.S. regulators may assert jurisdiction over and prosecute non-U.S. persons or entities if the trading activity has a connection to U.S. commerce.16
1Michael S. Selig, The Next Phase of Project Crypto: Unleashing Innovation for the New Frontier of Finance, Remarks of Chairman Michael S. Selig at CFTC-SEC Event on Harmonization (Jan. 29, 2026), CFTC (speech), https://www.cftc.gov/PressRoom/SpeechesTestimony/opaselig1. 2See 17 CFR § 38.200. 3See Michael S. Selig, The Next Phase of Project Crypto: Unleashing Innovation for the New Frontier of Finance, Remarks of Chairman Michael S. Selig at CFTC-SEC Event on Harmonization (Jan. 29, 2026), CFTC (speech), https://www.cftc.gov/PressRoom/SpeechesTestimony/opaselig1; see also 89 Fed. Reg. 48968 (June 10, 2024), https://www.federalregister.gov/documents/2024/06/10/2024-12125/event-contracts. 4See U.S. Commodity Futures Trading Comm’n, CFTC Staff Advisory Letter – Withdrawal, No. 26-04 (Feb. 4, 2026). The CFTC was unable to finalize the 2024 proposed rule before the transition, and the proposal has now reached a predictable dead end. In remarks delivered at the Joint SEC-CFTC Harmonization Event, Chairman Selig announced that the Commission will not proceed with the 2024 proposal, signaling a recalibration of priorities in the prediction markets space. This shift appears to reflect concerns that the 2024 proposal (1) clashed head-on with constitutional principles; (2) adopted a broad definition of “gaming”; (3) categorically classified event contracts as contrary to the public interest, regardless of their specific terms and conditions; (4) relied on an antiquated “economic purpose test”; and (5) reflected a misunderstanding of the distinction between federal derivatives regulation and state gaming regulation. 5See U.S. Commodity Futures Trading Comm’n, CFTC Staff Advisory Letter No. 25-36 (Sept. 30, 2025). 6Several recent cases highlight the divergent approaches emerging in different jurisdictions. 7See U.S. Commodity Futures Trading Comm’n, CFTC Staff Advisory Letter No. 25-36 (Sept. 30, 2025). 8See 7 U.S.C. § 7a-2(c)(5)(A). 9See 7 U.S.C. § 1a(19)(iv), 7a-2(c)(5)(A). 10CNBC, New York Attorney General Warns Against Prediction Market Trades Ahead of Super Bowl (Feb. 2, 2026), https://www.cnbc.com/2026/02/02/new-york-ag-prediction-markets-super-bowl-warning.html. 11Id. CFTC-regulated markets have extensive customer protections, including segregation of customer funds and limitations on permitted depositories for such funds. 12Id. 13Jessica Corso, SDNY Chief Says Office Has Eye on Prediction Markets, Law360 (Feb 6, 2026), https://www.law360.com/sports-and-betting/articles/2438607?nl_pk=66c6660d-02d0-4c8d-88b7-e422aeb64d17&utm_source=newsletter&utm_medium=email&utm_campaign=sports-and-betting&utm_content=2026-02-06&read_main=1&nlsidx=0&nlaidx=0. 14Id. 15“[P]rincipal place of business means the location from which the officers, partners, or managers of the legal person primarily direct, control, and coordinate the activities of the legal person.” Commodity Futures Trading Comm’n, CFTC Letter No. 25-14, Staff Interpretation Regarding Certain Cross-Border Definitions (May 21, 2025). 16See Commodity Futures Trading Commission, CFTC Charges Commodity Pool Operators and Their Co-Chief Investment Officer with Deception and Manipulation in Connection with Swaps and Supervision Failures, CFTC Release No. 8640-22 (Dec. 15, 2022), https://www.cftc.gov/PressRoom/PressReleases/8640-22 (announcing charges against Neil Phillips, a resident of the United Kingdom, “with engaging in a deceptive and manipulative scheme to illegally trigger payouts on two large binary option contracts”); see also Commodity Futures Trading Commission, CFTC Charges Binance and Its Founder, Changpeng Zhao, with Willful Evasion of Federal Law and Operating an Illegal Digital Asset Derivatives Exchange, CFTC Release No. 8680-23 (March 27, 2023), https://www.cftc.gov/PressRoom/PressReleases/8680-23 (announcing “a civil enforcement action in the U.S. District Court for the Northern District of Illinois charging Changpeng Zhao and three entities that operate the Binance platform with numerous violations of the Commodity Exchange Act and CFTC regulations”); see also Commodity Futures Trading Commission, CFTC Issues Order Against Crypto Prime Brokerage Firm for Unlawfully Providing U.S. Customers Access to Digital Asset Derivatives Trading Platforms, CFTC Release No. 8909-24 (May 13, 2024), https://www.cftc.gov/PressRoom/PressReleases/8909-24 (describing the “filing and settling charges against Falcon Labs, Ltd., an entity organized under the laws of the Seychelles, for failing to register with the CFTC as a futures commission merchant”).
New facility will enable end‑to‑end discovery across key disease areas and technology platforms, joining existing global research sites including Cambridge, Massachusetts, and Basel, Switzerland
Approximately 466,000-square-foot site withAI-enabled discovery capabilities, expected to house about 1,000 employees
Part of a USD 23 billion US investment to expand R&D and advanced manufacturing to reach more patients
Basel, February 6, 2026 – Novartis today broke ground on a new, state-of-the-art global Biomedical Research center in San Diego, California, designed to provide world-class scientific infrastructure and drug-discovery capabilities to advance the company’s pipeline for patients.
Once operational in 2029, the approximately 466,000-square-foot facility is expected to house about 1,000 Novartis employees and integrate seamlessly with Novartis global research sites including Cambridge, Massachusetts, and Basel, Switzerland, enabling integrated discovery efforts across regions.
“This new research center will strengthen our scientific leadership and accelerate the discovery of transformative medicines for patients worldwide, while deepening our connectivity with biotech, academic and technology partners across the region,” said Fiona Marshall, President of Biomedical Research at Novartis. “Designed to power future drug discovery, with a focus on genetics and human biology in key therapeutic areas such as neuroscience and oncology, it will create a single Novartis research center within one of the world’s premier life sciences ecosystems—accelerating our pipeline from discovery to patients.”
Government leaders and community partners joined Novartis employees in San Diego to mark the start of construction at the groundbreaking ceremony.
“This investment by Novartis reinforces San Diego as a place where breakthrough science happens and where innovation translates into high-quality jobs and life-changing medicines,” said Todd Gloria, San Diego Mayor. “San Diego is a global leader in life sciences because we bring together world-class talent, cutting-edge research and a collaborative ecosystem that turns discovery into impact. We’re proud to welcome this new research center and to continue building an economy rooted in innovation and results.”
The new facility builds on more than 25 years of Novartis research and development in San Diego and will support end-to-end drug discovery across core disease areas, including neuroscience, global health, oncology and age-related diseases and regenerative medicine. It will also expand the company’s capabilities in next-generation technology platforms, such as gene and cell therapies, RNA-based therapies, biologics and targeted protein degradation, and will advance novel delivery systems that open new therapeutic possibilities.
“Rooted in a strong legacy of innovation in San Diego and California, we are inspired to shape the future—driving new breakthroughs that will transform care for millions of patients worldwide,” said Thierry Diagana, California Sites Head and Global Head of Global Health, Biomedical Research, Novartis.
With advanced artificial intelligence, data and computational capabilities embedded throughout, the facility will be a key part of the global Novartis Biomedical Research network, helping share insights, scale innovation, and deliver meaningful breakthroughs for patients worldwide.
The San Diego research facility is a key pillar of the company’s USD 23 billion US investment. This includes a new flagship manufacturing hub in North Carolina; the opening of a radioligand therapy (RLT) manufacturing facility in Carlsbad, California; investments to expand existing facilities in Indiana and New Jersey; and plans to build new RLT manufacturing facilities in Florida and Texas.
Disclaimer
This press release contains forward-looking statements within the meaning of the United States Private Securities Litigation Reform Act of 1995. Forward-looking statements can generally be identified by words such as “potential,” “can,” “will,” “plan,” “may,” “could,” “would,” “expect,” “anticipate,” “look forward,” “believe,” “committed,” “investigational,” “pipeline,” “launch,” or similar terms, or by express or implied discussions regarding potential marketing approvals, new indications or labeling for the investigational or approved products described in this press release; or regarding potential future revenues from such products; or regarding discussions of strategy, plans, expectations or intentions, including discussions regarding our continued investment into new US R&D or manufacturing capabilities. You should not place undue reliance on these statements. Such forward-looking statements are based on our current beliefs and expectations regarding future events, and are subject to significant known and unknown risks and uncertainties. Should one or more of these risks or uncertainties materialize, or should underlying assumptions prove incorrect, actual results may vary materially from those set forth in the forward-looking statements. There can be no guarantee that the investigational or approved products described in this press release will be submitted or approved for sale or for any additional indications or labeling in any market, or at any particular time. Nor can there be any guarantee that such products will be commercially successful in the future. Neither can there be any guarantee that the expected benefits from the plans and investments described in this press release will be achieved in the expected timeframe, or at all. In particular, our expectations regarding such products could be affected by, among other things, the uncertainties inherent in research and development, including clinical trial results and additional analysis of existing clinical data; regulatory actions or delays or government regulation generally; global trends toward health care cost containment, including government, payor and general public pricing and reimbursement pressures and requirements for increased pricing transparency; our ability to obtain or maintain proprietary intellectual property protection; the particular prescribing preferences of physicians and patients; general political, economic and business conditions, including the effects of and efforts to mitigate pandemic diseases; safety, quality, data integrity or manufacturing issues; potential or actual data security and data privacy breaches, or disruptions of our information technology systems, and other risks and factors referred to in Novartis AG’s current Form 20-F on file with the US Securities and Exchange Commission. Novartis is providing the information in this press release as of this date and does not undertake any obligation to update any forward-looking statements contained in this press release as a result of new information, future events or otherwise.
About Novartis Novartis is an innovative medicines company. Every day, we work to reimagine medicine to improve and extend people’s lives so that patients, healthcare professionals and societies are empowered in the face of serious disease. Our medicines reach more than 300 million people worldwide.
Reimagine medicine with us: Visit us at https://www.novartis.com and connect with us on LinkedIn, Facebook, X/Twitter and Instagram.
The International Atomic Energy Agency (IAEA) and OCP Group, a global leader in plant nutrition solutions, have launched a five-year strategic partnership to accelerate scientific innovation for sustainable agriculture and resilient food systems. This collaboration will directly support the Atoms4Food initiative.
Under the agreement, the IAEA and OCP Group will launch a coordinated research project (CRP) applying nuclear and isotopic techniques to improve fertilizer efficiency, enhance crop nutritional quality and reinforce the sustainability of food systems. The project is designed to deliver practical, on‑the‑ground benefits for farmers, particularly in regions facing acute food security challenges.
“This partnership with OCP represents an important advancement in how strategic collaboration can amplify the impact of our Atoms4Food initiative”, IAEA Director General Rafael Mariano Grossi stated. “OCP’s significant commitment and on-the-ground expertise, combined with the IAEA’s unique expertise in nuclear techniques, will translate advanced science into practical solutions for farmers. Together, we will generate the evidence and tools needed to use fertilizers more efficiently, cultivate more nutritious crops and strengthen climate-resilient food systems, particularly in the regions that need them most.”
Research will focus on optimizing the management of key macronutrients such as nitrogen and phosphorus, as well as essential micronutrients including zinc, iron and selenium. Using isotopic techniques, the project will generate robust data to support the “4Rs” of nutrient stewardship – using the right source, at the right rate, at the right time and in the right place – providing farmers with actionable, evidence-based guidance.
“This collaboration marks a strategic milestone for OCP Group and a major step forward in our mission to strengthen global food security”, stated Meriem El Asraoui, Chief Global Affairs Officer of OCP Group. “By combining the IAEA’s world-class expertise with OCP Group’s deep experience and groundbreaking innovations in plant and soil nutrition, we will generate transformative knowledge, support researchers and farmers on the ground and advance agricultural practices that deliver higher yields, better nutrition and long‑term environmental stewardship.”
Benefits for IAEA Member States
This partnership will generate high-quality data to inform public policy, guide fertilizer innovation and support the transition to climate- and nature-positive agriculture. It will also reinforce scientific cooperation between Africa, the IAEA and global research networks, helping countries adopt nutrient stewardship best practices that improve soil health and crop yields at scale, directly contributing to global food security.
About the IAEA and Atoms4Food
The IAEA serves as the world’s foremost intergovernmental forum for scientific and technical co-operation in the peaceful use of nuclear technology. Established as an autonomous organization under the United Nations (UN) in 1957, the IAEA carries out programmes to maximize the useful contribution of nuclear technology to society while verifying its peaceful use.
The Atoms4Food initiative, jointly launched by the IAEA and the Food and Agriculture Organisation of the United Nations (FAO) in 2023, helps countries boost food security and to tackle growing hunger. The initiative seeks to provide countries with ground-breaking solutions tailored to their specific needs and circumstances by harnessing the advantages of nuclear techniques along with other advanced technologies to enhance agricultural and livestock productivity and natural resources management, reduce food losses, ensure food safety, improve nutrition and adapt to the challenges of climate change.
For more information, visit www.iaea.org.
To keep abreast of the IAEA’s latest developments, follow the IAEA on Facebook, Instagram, LinkedIn, X and Weibo.
About OCP Group
OCP Group is the trusted custodian of one of the most vital natural resources – phosphate – and a global leader in plant nutrition and phosphate-based solutions. Founded in 1920, headquartered in Morocco and anchored in Africa, the Group operates globally with more than 17,000 employees and over 350 customers across five continents.
We harness the full potential of phosphate to produce customized fertilizers, and deploy competitive industrial products and solutions, transforming a natural resource into a source of economic and strategic value.
Our US$13 billion Investment Plan (2023-2027) is designed to help us produce cost-competitive, low carbon electricity, water, ammonia and hydrogen to power our operations, while advancing our ambition to spearhead a decarbonized and circular industrial model.
The Group’s innovation ecosystem – catalyzed by UM6P and INNOVX – drives cutting-edge research, talent development and industrial ventures across Africa and beyond. Our strategic entities, OCP Nutricrops and OCP SPS (Specialty Products & Solutions), and our strategic business unit, Rock Solutions, maximize the value of Morocco’s phosphate reserves while creating shared prosperity.
By bringing phosphorus to life, OCP Group is turning custodianship into a model of competitive and sustainable industrial transformation – delivering resilient growth for Morocco, Africa and the world.
Noncommunicable diseases (NCDs) are the most common cause of death worldwide []. In Malaysia [], NCDs particularly impact lower-income households []. Therefore, health surveillance in this population is required to better understand policy interventions that may improve health outcomes in Malaysia. Dietary risk factors accounted for 10% of all deaths globally in 2021 []; therefore, measuring eating is a crucial component of health surveillance. Traditional methods for measuring eating and dietary intake include food diaries, 24-hour recalls, diet histories, or food frequency questionnaires. While methods relying on memory of past behavior are subject to error like recall bias, prospective methods like diaries are affected by reactivity, where real or reported behavior is altered owing to the process of documenting food intake in real time. Underreporting is common in all existing methods, with an estimated 263 kcal per day typically missing from self-reported intakes compared with objective measures []. Underreporting varies with food type and eating occasion, with snacks and snack foods more likely to be left out of a self-reported record [-]. Online tool, such as Intake24 (Newcastle University), that guide users through a 24-hour recall process aim to reduce researcher burden in coding data collected. Photographic methods, where participants are asked to take pictures of their meals rather than write down each food and drink along with its portion size, aim to offer a more objective approach to add portion size estimation and reduced participant burden for capturing real-time food intake [-]. However, moving 24-hour recall online has not yet altered estimated underreporting [,], and issues with remembering to take photos before consuming foods as well as automating the estimation of foods and nutrients [] in photos mean that outstanding challenges in dietary assessment methods remain []. Therefore, the feasibility of using alternative methods needs to be considered.
Ecological momentary assessment (EMA) is the repeated sampling of current behaviors in real-time in a natural environment []. EMA has evolved to be primarily delivered using mobile phones (mEMA), which have improved response rates compared with original pen and paper methods [-]. There is a large volume of literature on EMA using smartphones (n=796 studies) []. While diet is the second most commonly studied topic, it still only accounts for 4% (35/796) of these studies. Studies of diet using EMA in young people are primarily in the United States and Europe, with just 2 studies in Asia, in China, and Taiwan [,].
Liao [] highlighted that response rates and compliance with EMA protocols were rarely reported. Since then, reporting of compliance has improved, but the response latency remains unknown from many studies []. Response rates to mEMA of diet are a median of 74% [], which is similar to mEMA of all topics (mean 75% (IQR 64%-84%) []. A review of mEMA for diet in young people (16-30 years) showed response rates mostly exceeded 80% [], whereas at younger ages poorer responses <80% are more often observed []. Lower response rates have also been associated with weekends versus weekdays [], when participants receive more prompts during the day [,], and in males versus females [].
Smartwatches are an emerging technology for collecting data alongside sensor data using micro-EMA (μEMA) protocols. This captures information using single-tap responses to brief questions, which is suitable for the small screens on these devices [,]. In adults, μEMA has been found to yield higher compliance rates despite more frequent sampling than mEMA and is perceived by users as less distracting []. Further, the use of μEMA significantly improves response rate (mean 72% vs 82%) but remains rare, with only 12 studies on any topic in any age group []. Despite these advantages, some limitations of μEMA have been reported in the literature, including limited battery life and technical problems such as problems with charging [].
In children and adolescents, the use of pen and paper EMA to measure diet has typically been implemented outside of school hours []. Internet-connected devices such as mobile phones are often used for mEMA data collection []. These may be less suitable for child and adolescent populations, where 40% of education systems now ban the use of smartphones in school []. Further, devices such as smartwatches that can function without an internet connection may be better suited to rural, semirural, and low-resource settings where communication infrastructure may be less well-developed []. Therefore, smartwatches offer the potential to implement EMA across the whole day, with the potential for additional advantages such as improved compliance and response rates [,]. To our knowledge, only two diet studies involving adults in the United Kingdom and the United States have reported on EMA with smartwatches, and none have involved children [,,].
Therefore, this study investigates the feasibility of using smartwatch-based μEMA to record eating patterns in Malaysian children and adolescents. The collected μEMA data are used to examine the completeness of the collected data and factors associated with response rates, alongside survey responses assessing participants’ experience during the study. Establishing the feasibility of this novel dietary measurement tool is an important first step to inform utility and any required refinement prior to deployment for dietary measurement more widely.
Methods
SEACO-CH20 Study
The South-East Asian Community Observatory (SEACO) health and demographic surveillance cohort is a dynamic cohort of 13,335 households in Segamat, a semirural region in the state of Johor Darul Takzim, Malaysia. The cohort was established in 2012, with surveys, blood tests, and physical measurement data collected from participants. In 2013 and 2018, health surveys were conducted on ~25,000 adults and children, 25,168 in 2013 and 24,710 in 2018.
All households (18,602) in 5 subdistricts, which SEACO operates, were invited to participate in the 2017 census. Altogether 11,617 households (40,184 residents) accepted our invitation. In 2018, participants who were involved in the 2017 census and were older than 5 years were invited to participate in the 2018 health round data survey, to which a total of 24,710 participants agreed. Potential participants from 3 subdistricts were preselected and approached via telephone using the existing health database (HR 2028). Participants’ parents were approached via telephone for recruitment before the home visit.
Children and adolescents aged 7-18 years who were part of the SEACO cohort were invited to participate in the SEACO Child Health 2020 update (SEACO-CH20) study; a systematic review of EMA studies in youth recommended 7 as a lower age limit for EMA []. The eligibility of households was limited by location due to the safety measures implemented during the COVID-19 pandemic to reduce the risk to participants, households, and fieldworkers. Therefore, the 1993 children and adolescents invited to participate were from only 3 of the 5 SEACO subdistricts (Jabi, Sungai Segamat, and Gemereh) in the Segamat district.
Data collection visits to individual households were performed in person from November 1, 2021, to July 31, 2022. The data were collected as part of a larger study and included surveys, physical measurements, such as height, weight, blood pressure, waist and hip circumference, and blood sample collection. Participants were given wrist-worn Axivity AX6 6-axis accelerometers to monitor their physical activity, which were worn 24 hours per day over 7 days []. A random subset of the participants were also given TicWatch C2 (Mobvo) Android smartwatches to record eating and drinking with μEMA as part of this feasibility study, using a smartwatch μEMA system developed within the research team. The smartwatch system used was an adaptation of a µEMA system first used in a study involving high-temporal density longitudinal measurement of alcohol use [] by a subset of the research team who developed this system. Participants wore these devices over the same 7-day period as the accelerometers. As they are both wrist-worn devices, this may have affected the acceptability of the smartwatch. Participants were briefed on the use of these devices by the data collectors, including how to charge them and how to replace them after charging. The original data collection plan can be seen in Figure S1 in . A completed Checklist for Reporting Ecological Momentary Assessment Studies (CREMAS) [] can be found in .
Study Participants
Participants for the SEACO-CH20 study were selected from the larger SEACO cohort. Parents of participants provided consent for their children to participate in SEACO-CH20. A random subsample of 100 participants were each invited to wear a smartwatch.
Data Collection
SEACO-CH20 fieldworkers performed 2 home visits to collect the data. The smartwatch was distributed on the initial visit, and the participants were briefed on how to use and charge it. They were instructed to wear the device for the next 8 days, on “the wrist that [they] use to write.” The smartwatches were collected during the second home visit, and the participants were asked to complete a questionnaire on their experience with the devices. Questionnaires assessed the participants’ attitudes toward several aspects of the smartwatch study, including ease of use, their attitude toward charging, and their overall experience. Since children as young as 10 were asked to complete the questionnaires, a reduced set of acceptance questions was used to reduce burden. This was based on similar pilot work using novel methods in the ALSPAC G2 study [] and included the following questions:
Overall, how would you rate your experience of using the smartwatch during the study, on a scale from 1 (didn’t like it at all) to 5 (really liked it)?
If you were asked to use the smartwatch again in another study, would you participate?
How many days in total did you wear the smartwatch for?
If you wore the smartwatch for less than 8 days, what were the main reasons for not wearing it longer?
Parents of participants aged 7-9 years completed the survey on behalf of their children, while participants aged 10 years and older filled out the survey themselves. The full text of the survey can be found in Figure S2 in .
Smartwatch μEMA Questions
During the study, the smartwatch prompted participants once every hour to enter any food or drink that they had consumed in the last hour. These prompts were scheduled to appear once every hour from 9 AM to 8 PM, so participants were expected to interact with the smartwatch 12 times throughout the day. We chose this hourly prompt frequency to maximize the chance that eating and drinking events were less likely to be missed and to capture more fine-grained temporal patterns in eating behavior. As this was a feasibility study, this choice was justified, given that μEMA has been shown to improve compliance despite more prompts than mEMA in adults []. The smartwatch interface included the following 5 questions that the participants completed for each item consumed:
Have you had any food or drink in the last hour? Options: yes, no
What did you have? Options: meal, snack, drink
What size was it? Options: small, medium, large
What did you use to eat? Options: hands, fork/spoon, chopsticks
Where were you? Options: home, school, elsewhere
A possible future use of this methodology, once fully refined, could include automated eating detection []. Therefore, we included a question on the type of cutlery used (“What did you use to eat?”), as lack of information on utensil type has been highlighted as a limitation of some datasets used for algorithm development related to automated eating detection [].
After entering this information for one item consumed, they were asked, “Any more food or drink to record?” and could then start again to add another entry. Therefore, each consumption entry either indicates that the participant did not eat or drink in the last hour or contains the answers to the above questions for a particular meal, drink, or snack, linked to an hour period within a day. If participants ignored the prompts, they would receive a reminder prompt after 1 minute; if they continued to ignore the prompt for a further 1 minute, the prompt would disappear and “no response” would be recorded by the smartwatch.
Participants could choose “back” on each question screen to return to a previous question and update their response. However, after submitting their answers for a particular item (ie, completing the “where were you” question for that item), they would not be able to return to that entry.
An additional prompt (the “catch-up”) was scheduled every morning at 8 AM asking if they had consumed any food or drink on the previous day that had not been recorded on the smartwatch. If they indicated “yes,” they were asked the same questions as above. Catch-up entries did not have an associated eating time but were labeled as catch-up–type events, indicating that they applied to the previous day.
The smartwatch study was co-created and piloted with the Malaysian research team and the English was translated into Malay for use on the smartwatch. All the original data collection was in Malay. The smartwatch protocol, including the prompts and possible responses, can be seen in Tables S1-S3 in .
Smartwatch Data Cleaning
Smartwatches were distributed by fieldworkers partway through the day, and μEMA responses on this distribution day were removed from analyses. The study period is taken to be the subsequent 7 days after this distribution day.
The version of the EMA software we used did not save the hour period to which each entry belonged. Therefore, we needed to infer this from the submission timestamp, the date and time a particular entry was submitted. As entries for the same hour period are submitted one after the other, we used a time window to group nearby entries into a single “eating event,” which is intended to capture the participants’ responses to one prompt. A 30-minute window was chosen to group nearby prompts, as we expect this to collect entries from the same eating event without grouping prompts from adjacent hours. Previous work from diet diaries suggests that 30 minutes is a reasonable cut-point to distinguish independent eating occasions []. Occasionally, there may be participants with more than 12 eating events per day, for example, if they took more than 30 minutes to finish responding to a prompt. This occurred on 26 occasions, less than 5% of the total 574 (82 participants multiplied by 7 days) study days.
The μEMA data used was restricted to the 7 days after the distribution day. During the data review, we identified an issue with the collected data where there were sometimes multiple identical entries for a given intake event due to an issue with the μEMA software. Therefore, duplicate entries were identified as any pair of entries with identical contents (same meal type, portion size, utensil, and location), for the same hour period, and entered within 5 minutes of each other. The first such entry was kept in each case. Around 588 duplicate responses were removed of 10,539. Data cleaning was performed in Python (version 3.10.0; Python Software Foundation).
Response Rate
The response rate was calculated as the proportion of prompts responded to (with either at least one item consumed or an entry stating that they did not eat or drink anything in the previous hour). The response rate tells us the extent that participants engaged with the smartwatch app throughout the day but not the extent that the data entered are complete, that is, whether all intake events were recorded. We therefore summarized the number of each type of meal entry (meal, drink, snack, or no food or drink) submitted per day across our sample. For these summaries we included only participants who took part in the study outside the Ramadan fasting period (April 3, 2022-May 1, 2022; n=67), since fasting participants are likely to enter fewer eating events during the day. The mean of participants’ response rates per day was recorded, and the median and quartiles of these were reported.
Attrition from the study was examined by identifying the last day each participant responded to any smartwatch prompt; participants who had ceased responding to the prompts are referred to as “inactive.”
Statistical Analyses
We summarized response rates to each individual prompt of the smartwatch μEMA and the participants’ experiences based on the survey questions. We used mean for continuous variables, n (%) for categorical variables, and median and IQR for ordinal or nonnormally distributed continuous variables.
We used a mixed-effect logistic regression model for the response (yes or no) to an individual prompt on a specific study day of data collection for each participant. A fixed effect term was included for study day (from the first to the seventh day) as a continuous linear trend. The time of day was also included as a fixed effect to capture nonlinearity in response throughout the day, grouping the prompts by the nearest hour as follows to decrease the number of parameters in the model:
Morning (9-11 AM)
Lunchtime (12-2 PM)
Afternoon (3-5 PM)
Evening (6-8 PM)
Random intercept and random slope terms were included for study day within each participant. Estimates are provided as odds ratios (OR) and 95% CIs, interpreted as the multiplicative change in the odds of a participant responding to an individual prompt. The degree of difference between participants was summarized in the intraclass correlation coefficient.
To evaluate if changes in participation across wear days differed in boys versus girls, we repeated this base model, adding a fixed term for sex and an interaction term between sex and study day. Similarly, we explored differences by age group (Malaysian primary school age: 7-12 years versus secondary school age: 13-18 years) by adding a fixed term for age group and an interaction term between age group and study day to the base model.
Analyses were performed in Python version 3.10.0 and R (version 4.2.2; R Core Team) []. All of our analysis code is publicly available []. Git tag 3.0 (The Git Project) corresponds to the version of the analyses presented here.
Ethical Considerations
Written informed consent was obtained from parents or guardians on behalf of the participants. Children and adolescents were also asked to provide their written assent to participate in the study. Ethical approval was obtained from the Monash University Human Research Ethics Committee on March 17, 2020 (Project ID: 23271) and the University of Bristol REC Case no. 2020–4208 (ID nr: 1304255) prior to any data collection. The study was conducted in accordance with the Declaration of Helsinki for experiments involving humans.
Participants were given a token worth up to RM25 to compensate for their time participating in the study. This was divided into RM5 for completion of each of the following components: (1) questionnaires, (2) health check, (3) blood sample, (4) activity monitor, and (5) smartwatch. They were also given a free health screen and a direct referral to the government primary health care clinic if they were identified as high risk.
Study data have been deidentified and can be freely requested from SEACO, Monash University Malaysia Institutional Data Access at “mum.seaco@monash.edu” for researchers meeting the criteria for access to confidential data. Please refer to the web resource hosted on Monash University’s website [] for more information.
Results
Participants
A flowchart showing the study participants can be seen in . Parents of 728 participants consented to their children’s participation in SEACO-CH20, of which 626 provided demographic (age, sex, and ethnicity) and accelerometer data for the larger study. Of these, 100 participants were randomly invited to wear a smartwatch for this smaller feasibility study. Of the 100 participants invited to participate in the smartwatch substudy, 83 participants agreed. The reasons for nonparticipation included concern the device was not comfortable (n=3) and allergies (n=2). The remaining participants rejected the smartwatch without comment. One further participant accepted the smartwatch study but removed the EMA app from the watch during the study period, rendering their data unrecoverable, resulting in 82 participants who provided smartwatch data.
The sex, ethnicity, and age breakdown for all participants who took part in the smartwatch study can be seen in .
Figure 1. Study flowchart. Eligible participants were selected from the SEACO Health Round Survey 2018 (SEACO HR-2018) cohort. Reasons for rejecting the smartwatch study included concern about discomfort and allergies. SEACO-CH20: South-East Asian Community Observatory Child Health 2020; SEACO HR-2018: South-East Asian Community Observatory Health Round Survey 2018;.
Table 1. Summary of participant demographics (N=83).
Participant characteristic
Smartwatch participants, n (%)
Sex
Female
53 (64)
Male
30 (36)
Ethnicity
Malay
73 (88)
Non-Malay
10 (12)
Age (years)
7-12
24 (29)
13-15
28 (34)
16-18
31 (37)
Smartwatch Responses
The median prompt response rate was 69% (IQR 52%-82%).
The number of participants who became inactive on each day can be seen in . The majority (55/82, 67%) of participants were active until day 7, that is, they responded to at least 1 prompt on day 7.
Figure 2. The number of participants who became inactive on each day (N=82), that is, having responded to no μEMA prompts after this day. All participants were active for at least one day.
The median and IQR in the number of entries of each type made by each participant per day are summarized in . Fifteen participants took part (at least partially) during Ramadan and so were excluded from these summaries. Only a minority of intake events were submitted as catch-up entries (N=125 catch-up entries versus 4705 noncatch-up).
Table 2. The median and IQR of the number of noncatch-up entries per day per participant, for participants whose study period did not intersect with Ramadan (N=67).
Study day
Meal, median (IQR)
Drink, median (IQR)
Snack, median (IQR)
1
2 (2-4)
3 (1-5)
1 (0-2)
2
2 (1-3)
2 (1-5)
1 (0-2)
3
2 (1-3)
2 (0-3)
0 (0-1)
4
2 (1-3)
1 (0-3)
1 (0-1)
5
1 (0-2)
1 (0-3)
0 (0-1)
6
1 (0-2)
1 (0-2)
0 (0-1)
7
1 (0-2)
0 (0-2)
0 (0-1)
Response Rate Across and Within Study Days
The response rate for individual prompts had a median (IQR) of 67% (50-83). The response rate on each day ranged from 83% (66-92) on day 1 to 58% (33-75) on day 7.
shows the response rate with study day and time. The response rate decreased across study days (OR for each additional day of the study: 0.73 (95% CI 0.64-0.83). The response rate was lowest at the beginning of the day; the OR and 95% CIs are summarized in . The intraclass correlation coefficient was 0.207, which indicates that approximately 21% of the total variance in prompt response behavior was attributable to between-participant differences.
Figure 3. Participants’ response rates against study day (left) and time of day (right; N=82).
Table 3. The odds ratios for responses at different times of day, taking breakfast time (9-11AM) as the reference level. The response rate was lowest at the beginning of the day.
Time
Odds ratio (95% CI)
Breakfast (9-11 AM)
1a (0-0)
Lunchtime (12-2 PM)
1.69 (1.43-2.12)
Afternoon (3-5 PM)
1.49 (1.25-1.76)
Evening (6-8 PM)
1.27 (1.07-1.51)
aReference level.
The results of analyses estimating differences due to sex and age are shown in . Girls responded more often to the μEMA prompts compared with boys (OR 1.71, 95% CI 1.03-2.84). However, the daily patterns were similar for both sexes (interaction term OR 1.07, 95% CI 0.93-1.23). Response rate did not differ between age groups (OR 0.73, 95% CI 0.42-1.27), and daily response patterns were similar for the 2 age groups (study day-by-age interaction OR 1.11, 95% CI 0.95-1.29).
Figure 4. Association of micro-interaction ecological momentary assessment (μEMA) prompt response with study day stratified by sex (left) and age (right).
Evaluating Participants’ Survey Responses on Acceptability
A summary of responses to questions about smartwatch acceptability and wear time is provided in .
A total of 54 out of 83 (65%) participants rated their experience using the smartwatch positively (a rating of 4 or 5 out of 5), 20 out of 83 (24%) gave a neutral rating (3 out of 5), and 8 out of 83 (10%) rated it negatively (1 or 2 out of 5). In addition, 27 out of 83 (33%) participants said that they would be happy to participate in future studies using the smartwatch, while 39 out of 83 (47%) said maybe, and 12 out of 83 (14%) said no. The majority of participants who responded (61/83, 73.5%) reported wearing the smartwatch for the entire duration of the study (8 days).
Table 4. The survey questions. Participants were directed to wear the smartwatch on the day that the watch was distributed and for the 7 subsequent days, making 8 days total. Missing data and participants who refused to respond are not included (N=83).
Question and response
Participants, n (%)
Overall, how would you rate your experience of using the smartwatch during the study, on a scale from 1 (didn’t like it at all) to 5 (really liked it)?
Negative (1 or 2)
8 (10)
Neutral (3)
20 (24)
Positive (4 or 5)
54 (65)
If you were asked to use the smartwatch again in another study, would you participate?
No
12 (14)
Maybe
39 (47)
Yes
27 (33)
How many days in total did you wear the smartwatch for?
5 or fewer
7 (9)
6
7 (8)
7
8 (10)
8
61 (73)
For those who reported that they did not wear the smartwatch for the entire duration of the study (22 participants), the most common reason was that they did not find it comfortable to wear (9/22, 41%). Other reasons included forgetting, that they did not see the benefit when they could not see the data, they were forbidden to wear it by school, and it ran out of battery (all 3 or fewer responses).
Summaries of the participants’ responses to the remaining survey questions can be found in Tables S4-S8 and Figure S3 in . Further summary statistics, including catch-up events and participants who took part during Ramadan, can be found in Tables S9-S11 in .
Discussion
Principal Findings
In this feasibility study of a smartwatch-based μEMA method to collect data on eating habits over 7 days in Malaysian children, we found that most participants (55/83, 67%) remained responsive to prompts up to the last day of the study. Participants were least likely to respond to prompts between 9 and 11 AM and most likely between 12 and 2 PM. The intraclass correlation coefficient was 20.7%, suggesting that while some variation in response pattern is attributable to between-participant differences, the majority of the variation (79.3%) was due to within-participant differences. The response rate dropped off day-on-day and was higher for female than male participants; no association was found between participant age group and response rate.
Our average response rate of 69% was lower than the average of 78% found in a meta-analysis of EMA in children and adolescents, including studies that prompted between 2 and 9 times daily []. That study found that prompting participants more often had a large negative effect on completion rate, which is further supported by Kraft et al [], which found a negative correlation of –0.12 between increased number of prompts and response rate (P=.009). Participants in our study were prompted 12 times a day, plus an additional catch-up prompt in the morning. We justified our original prompt frequency choice, as μEMA has been shown in adults to improve compliance despite a higher number of prompts than mEMA. However, in our study using a child and adolescent population, it is likely a higher prompt frequency may have had a negative effect on response rate, especially in the case of repeated “No food/drink” entries. It has been reported [] in nonclinical studies that “a higher average compliance rate was observed in studies that prompted participants 2-3 times daily (91.7%) compared with those that prompted participants more frequently (4-5 times: 77.4%; 6+ times: 75.0%).” This suggests that compliance may be improved by prompting participants less frequently, for example, by having 3 prompts daily at 11 AM, 3 PM, and 7 PM, although longer time intervals increase the reliance on memory, potentially affecting the completeness of recorded consumption events. It is also possible that our lower response rate might have been related to the number of questions asked in our μEMA protocol. A study that compared the deployment of 6 back-to-back multiple-choice questions delivered via a smartwatch versus a mobile phone found no difference in compliance between these 2 modalities. However, compliance was improved when single questions requiring a one-touch response were asked via a smartwatch, despite an increase in prompt frequency []. While on average, the participants (65%) rated the study protocol positively (either 4 or 5 out of 5), the response rate fell day-on-day. Participants were less likely to respond to prompts at the beginning or end of the day, compared with the middle of the day. Focus group or interview discussions were not feasible in this study due to COVID-19 restrictions but should be explored in future studies to determine the reasons for missing event prompts and nonresponses, which may include forgetting or being involved in a competing activity when the prompt is sent [].
Female participants had a higher response rate than male participants, consistent with previous findings [,]. There was little evidence of difference in the relationship between response and study day for male versus female participants. An analysis of the SEACO-CH20 accelerometer dataset [] found that a similar proportion of males and females had usable accelerometer data, suggesting that this was specific to the smartwatches rather than a difference with wrist-worn devices generally. Little evidence of a difference was found between response rates in the 7-12 years old and 13-18 years old age groups.
Although the subjective indicators suggested that most participants enjoyed wearing the smartwatch, only a minority of participants (27/83, 33%) indicated that they would be willing to participate in a similar study again; 39 out of 83 (47%) responded “Maybe.” Potential changes to the protocol that may improve compliance could include only wearing the smartwatch instead of both the smartwatch and accelerometer.
Strengths and Limitations
This is the first study exploring the feasibility of using smartwatch-based EMA in a population of children and adolescents from a low-to-middle-income country. This study was part of the SEACO study, using the SEACO-CH20 dataset, and lays the foundation for an improved understanding of the potential for wearable devices for measuring relationships between eating and cardiometabolic health. Data on 24-hour eating behaviors are important for informing future policy that may reduce cardiometabolic risk among children and adolescents and prevent progression to cardiometabolic disease in adulthood. While a food frequency questionnaire was completed as part of the larger SEACO-CH20 study and reported elsewhere [], this current study has assessed an alternative for recording behavior in real time.
However, this study did have some limitations. The number of questions in prompts may have affected compliance and should be discussed with participants to optimize the protocol for future studies in this population. It has been suggested [] that compliance can be improved by incentivizing participants with a monetary reward or raffle entries. Since this study concerns young people, one potential incentive method could be to gamify the EMA process using a level-up or promotion system in the app []. Previous studies have explored adding an end-of-day catch-up prompt, which has been found to improve the reporting of dinner []. Replacing the morning catch-up prompt with one in the evening may improve response rate, especially given that we found that participants were more likely to respond to prompts in the evening than in the morning. Future studies may additionally consider using the catch-up entries to impute missed entries on the previous day, which could give more complete data. Another suggestion could be to incorporate a short period of training to improve response rates, where responses are monitored in real time by researchers and participants are prompted directly by researchers if missing responses are common. Such an approach has been used to improve the accuracy of real-time food photography methods [].
The questionnaires used for acceptability and acceptance were not standard because we were motivated to use a reduced set of questions (that have been used in a previous publication []) to reduce the burden on the young participants. Future studies should consider if the emerging, more standard approaches to exploring acceptability and acceptance for wearable devices (eg, those based on Technology Acceptance Models) have been developed to the point at which varying levels of participant burden can be accommodated.
Unbalanced statistics limited our ability to assess differences across age, sex, and ethnic group. The larger SEACO-CH20 accelerometer study [] from which participants for this study were selected had more balanced statistics (49% female, 67% Malay, and 44% <13 years old), which suggests that the cohort used for smartwatch data may not represent the overall SEACO-CH20 cohort. In particular, we only had 4 participants aged 7-9 years, so further studies are required to better understand the feasibility of dietary μEMA in the younger participants. Participants in this study began wearing the devices on different days of the week, and it is possible that the day of the week could affect participation; for example, whether it is a weekend or weekday. A lower response rate on weekends has been previously documented by Battaglia []. We did not account for study start day due to the small sample size, and because schooling was disrupted throughout the study period due to the COVID-19 pandemic [].
An issue with entry duplication meant that some entries may have been removed that were actual events, not due to the software issue. This bug with the smartwatch software has since been fixed. Additionally, only the response time of the participant was recorded, and not the time that the prompt was sent. This means we had to infer which hour window the entries corresponded to; this could be programmed in the software.
Discomfort was the most common reason for nonwear cited by the participants, which may be unique to our study protocol that required participants to wear 2 wrist-worn devices on the same arm. Furthermore, the smartwatch used in our study was not specifically designed to fit smaller children. Efforts to make smartwatches less intrusive, for example, by making them smaller, may further improve response rate and study uptake.
Ramadan, a culturally important event in Malaysian society, which includes fasting in some population groups, took place during the course of the data collection period. This is likely to have affected the eating behaviors of participants who took part during this time. To ensure that the number of EMA entries was not influenced by Ramadan, we excluded participants from our analyses who wore the smartwatch during the Ramadan period, thereby further limiting our sample size.
Conclusions
This study extends previous eating behavior studies by exploring the use of μEMA in a population of children and adolescents in Malaysia and is the first such study to do so. Willingness to take part in the μEMA study was high, but poor response rates suggest that the number of questions asked per prompt or the high number of prompts per day may be too burdensome. While our smartwatch-based EMA app was largely based on the μEMA methods originally developed by Intille et al [], a key aspect of true μEMA implementation is the presentation of only one question at a time. In our approach, we chained questions to capture details on food and drink type, size, and consumption context, making it more accurately described as a modified μEMA. Further work is needed to explore different μEMA variations, including using fewer questions and/or fewer prompts, and identify devices that may be more comfortable for child and adolescent participants. The growing use of smartwatches amongst children, particularly in Southeast Asia may offer more opportunities for further study [].
For the purpose of open access, the authors have applied a Creative Commons Attribution (CC BY) license to any author accepted manuscript version arising from this submission.
The authors would like to express their appreciation to the SEACO Field Teams and survey participants. The SEACO [] funded the research detailed in this paper. The authors’ opinions, however, are their own, and there is no real or implied sponsorship from SEACO.
The Medical Research Council (MR/T018984/1) and the Ministry of Higher Education/UK-MY Joint Partnership on Non-Communicable Diseases (2019/MR/T018984/), both provided funding in support of this research. The SEACO health and demographic surveillance system is supported by Monash University. The funders had no involvement in the study design, data collection, analysis, interpretation, or the writing of the manuscript. Sophia M Brady is funded by the National Institute for Health and Care Research (NIHR) Applied Research Collaboration (ARC) North East and North Cumbria (NENC; NIHR200173). The NIHR Bristol Biomedical Research Centre funds Miranda EG Armstrong (NIHR203315). The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care.
Data cannot be shared publicly for confidentiality and ethical reasons. Deidentified data are available and can be freely requested from the South East Asia Community Observatory, Monash University Malaysia Institutional Data Access at “mum.seaco@monash.edu” for researchers who meet the criteria for access to confidential data. For more information, please refer to [].
None declared.
Edited by A Mavragani; submitted 01.May.2025; peer-reviewed by C Wang, P Delir Haghighi; comments to author 17.Aug.2025; revised version received 18.Nov.2025; accepted 30.Dec.2025; published 06.Feb.2026.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research (ISSN 1438-8871), is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.
