United States

Background
Artificial intelligence–enabled medical devices (AIMDs) are increasingly present in US clinical practice. However, nearly all AIMDs are cleared through the US Food and Drug Administration (FDA) 510(k) pathway, which does not require prospective human testing and thereby leads to many devices entering the market with limited or absent clinical validation. To investigate the clinical risk of this development and identify actionable areas for regulatory safeguards, we analyzed the contributory variables and clinical consequences of all AIMD recalls in the US.
Methods
The FDA AI/ML medical device database was used to compile a cohort of 950 FDA-cleared AIMDs. We manually extracted device information from all device approval summaries and identified 60 devices that were subject to recall and extracted data on these events using the FDA recall database. The S&amp;P Capital IQ NetAdvantage and PrivCo databases were used to analyze AIMD manufacturers based on company trading status, headquarters location, market capitalization, and acquisition history. Multivariable logistic regression (GraphPad Prism 10; 2-sided α = .05) modeled the odds of any recall, adjusting for validation status, manufacturer type, clearance year, and clinical specialty. 
Results
Of 950 FDA-cleared AIMDs, 60 (6.3%) were recalled across 182 total recall events. Of these 182 recalls, 11% have remained unresolved for &gt;3 years and 43% of recalls occurred within the first year following FDA clearance, double the reported rate for non-AIMD 510(k) devices. Diagnostic errors were the most common recall cause (109 recalls; 935,063 units), followed by functionality delay (44 recalls; 755,647 units), physical hazards (14 recalls; 8,192 units), and biochemical hazards (13 recalls; 76,257 units). We found that AIMDs without clinical validation had more recalls per device (mean 3.4 [SD 5.1]) than those with retrospective (1.9 [SD 1.5]) or prospective (2.0 [SD 1.4]) clinical testing. Similarly, larger publicly-traded companies manufactured 53.2% of AIMDs, but were responsible for 92% of recalls and 99% of recalled units. Multivariable analysis, revealed that lack of clinical validation (odds ratio [OR], 2.8; 95% CI, 1.6-4.7) and public company status (OR, 5.9; 95% CI, 2.4-14.6) were independently associated with higher likelihood of device recall. 
Conclusion
Recalls of FDA-cleared AIMDs are uncommon, but are concentrated early after clearance and predominantly involves products without clinical validation and manufactured by large publicly traded companies. Issuing time-limited clearances that lapse without confirmatory data, heightening pre-market clinical testing requirements, and increasing post-market surveillance measures are needed for improved identification and reduction of device errors.

AMA Research Challenge 2025 

From Expansion to Recalls: Artificial Intelligence-Enabled Medical Devices

Welcome to the largest national, multi-specialty research event **featuring research from 750 medical students, residents and fellows, and international medical graduates.**


#### Event Details
The 2022 American Medical Association Research Challenge includes a poster symposium hall and a semifinals hall with the top scored research. Log in at any time throughout the event to:
* **View posters and presentations** on a variety of topics and specialties
* **Score posters in the semifinals** to help decide the top five finalists (AMA member exclusive opportunity)
* **Connect with participants** and provide feedback to help advance their research

At the conclusion of this event, the AMA will announce the top five finalists who will advance to the AMA Research Challenge finals, present their research to an elite panel of judges and compete to win a $10,000 grand prize sponsored by Laurel Road.
<br>

#### Attend Live Forums
**Oct. 20, 6 — 7 p.m. Central**
<br>**How to plan a career in academic medicine** 
<br>Learn how to plan a career in academic medicine and get your questions answered. [Access Here](https://underline.io/events/363/lounge)
<br>
<br>**Speakers**
![](https://assets.underline.io/uploads/markdown_image/1/image/d5adb133dfc58eb0a6d469e9cedcd5d4.jpg)
Ricardo Correa, MD, EdD
![](https://assets.underline.io/uploads/markdown_image/1/image/5f9437ab87861538bf02b9ca7c0d926e.jpg)
Stuart M. Greenstein, MD
![](https://assets.underline.io/uploads/markdown_image/1/image/05b1b637c99286d558c371f8e8b185ce.jpg)
Stephanie Strohbeen, MD
<br>
<br>**Oct. 22, 11 a.m. — Noon Central**
<br>**How to get published**
<br>Learn tips to get your research published and get your questions answered. [Access Here](https://underline.io/events/363/lounge)
<br>
<br>**Speakers**
![](https://assets.underline.io/uploads/markdown_image/1/image/8ce6235e957d31cb5e6f9097cc6f6972.jpg)
Jessica Li, MD/PhD Candidate
![](https://assets.underline.io/uploads/markdown_image/1/image/9950ef2e0237f54cb3d8280334d42491.jpg)
Charles Lopresto, DO

### Sponsor
AMA would like to thank Laurel Road for sponsoring a $10,000 grand prize for the winner of the AMA Research Challenge. Laurel Road is not involved in the judging or scoring process to select the winner. The winner will be announced on Dec. 7 during the AMA Research Challenge finals. Payment of the grand prize will be directly from Laurel Road. See the AMA Research Challenge Official Rules for terms and conditions.

![](https://assets.underline.io/uploads/markdown_image/1/image/611e331d4a0eba0920f240cfc04b4d77.png)

<br>

**There is no registration fee for this event! **

Register to explore research in a variety of topics. AMA members get exclusive access to score posters in the semifinals.

**Please use the link below to register for free if you have not already done so:**

https://www.ama-assn.org/about/events/ama-research-challenge

Registration Is Required

AMA Research Challenge 2022

Welcome to the largest national, multi-specialty research event featuring research from 750 medical students, residents and fellows, and international medical graduates.


Background
Artificial intelligence–enabled medical devices (AIMDs) are increasingly present in US clinical practice. However, nearly all AIMDs are cleared through the US Food and Drug Administration (FDA) 510(k) pathway, which does not require prospective human testing and thereby leads to many devices entering the market with limited or absent clinical validation. To investigate the clinical risk of this development and identify actionable areas for regulatory safeguards, we analyzed the contributory variables and clinical consequences of all AIMD recalls in the US.
Methods
The FDA AI/ML medical device database was used to compile a cohort of 950 FDA-cleared AIMDs. We manually extracted device information from all device approval summaries and identified 60 devices that were subject to recall and extracted data on these events using the FDA recall database. The S&P Capital IQ NetAdvantage and PrivCo databases were used to analyze AIMD manufacturers based on company trading status, headquarters location, market capitalization, and acquisition history. Multivariable logistic regression (GraphPad Prism 10; 2-sided α = .05) modeled the odds of any recall, adjusting for validation status, manufacturer type, clearance year, and clinical specialty. 
Results
Of 950 FDA-cleared AIMDs, 60 (6.3%) were recalled across 182 total recall events. Of these 182 recalls, 11% have remained unresolved for >3 years and 43% of recalls occurred within the first year following FDA clearance, double the reported rate for non-AIMD 510(k) devices. Diagnostic errors were the most common recall cause (109 recalls; 935,063 units), followed by functionality delay (44 recalls; 755,647 units), physical hazards (14 recalls; 8,192 units), and biochemical hazards (13 recalls; 76,257 units). We found that AIMDs without clinical validation had more recalls per device (mean 3.4 [SD 5.1]) than those with retrospective (1.9 [SD 1.5]) or prospective (2.0 [SD 1.4]) clinical testing. Similarly, larger publicly-traded companies manufactured 53.2% of AIMDs, but were responsible for 92% of recalls and 99% of recalled units. Multivariable analysis, revealed that lack of clinical validation (odds ratio [OR], 2.8; 95% CI, 1.6-4.7) and public company status (OR, 5.9; 95% CI, 2.4-14.6) were independently associated with higher likelihood of device recall. 
Conclusion
Recalls of FDA-cleared AIMDs are uncommon, but are concentrated early after clearance and predominantly involves products without clinical validation and manufactured by large publicly traded companies. Issuing time-limited clearances that lapse without confirmatory data, heightening pre-market clinical testing requirements, and increasing post-market surveillance measures are needed for improved identification and reduction of device errors.

poster

Congratulations to all the researchers who have qualified for the AMA Research Challenge virtual poster symposium and semifinals!

Welcome to the **largest national, multi-specialty research event** featuring research from nearly 1,000 medical students, residents, fellows, and international medical graduates across six topic areas - basic science, clinical and translational research, clinical vignettes, health systems science, medical education, and public health and health policy.

#### Access the Event
Use the buttons below or navigation menu to access the event. 

[![](https://assets.underline.io/markdown_image/1/image/d4076f2146939317a29eb9f69c53ad41.png)](https://underline.io/events/506/sessions?searchGroup=all)
[![](https://assets.underline.io/markdown_image/1/image/6bcee0b523a4874c4292ca0415d88dde.png)](https://underline.io/events/506/posters?searchGroup=all)

#### Event Details
The 2025 AMA Research Challenge includes the following symposium halls:
* **Semifinals hall:** Features research from 54 medical students, residents and international medical graduates that received top scores during first round scoring. 
* **Poster symposium hall:** Features research from nearly 1,000 medical students, residents and international medical graduates categorized by topics and subtopics.

**Here’s how to participate:**
* **View posters and presentations** on a variety of topics and specialties
* **Score posters in the semifinals** to help decide the top five finalists (AMA member exclusive opportunity)
* **Connect with presenters** and provide feedback to help advance their research
* **New this year: Watch live poster presentations** by topic area and engage directly with researchers

At the conclusion of this event, the AMA announced the top five finalists who advanced to the AMA Research Challenge final round, presented their research to an elite panel of judges, and competed to win a $10,000 grand prize presented by Laurel Road.

**Addison Shenk, a fourth-year medical student from Edward Via Virginia College of Osteopathic Medicine, was the winner of the 2025 AMA Research Challenge and was awarded the grand prize of $10,000, presented by Laurel Road.**

#### A message from Dr. Bobby Mukkamala, AMA President
<div style="position: relative; width: 100%; padding-bottom: 56.25%; height: 0; overflow: hidden;">
  <iframe 
    src="https://underline.io/embed/136894-ama-video?t=0"
    title="AMA Welcome Video"
    allowfullscreen
    style="position: absolute; top: 0; left: 0; width: 100%; height: 100%; border: 0;">
  </iframe>
</div>


#### Recognized Institutions
In 2025, 1,400 abstracts were submitted across six topic areas representing 174 medical schools and more than 150 residency and health care institutions. [See if your institution is represented this year!](https://drive.google.com/file/d/1eFrTqCmCeobx9pr5drJK2hllvsIpFNOY/view?usp=drive_link)

#### Grand Prize Sponsor
AMA would like to thank Laurel Road, a brand of KeyBank, for sponsoring a $10,000 grand prize for the winner of the AMA Research Challenge. Laurel Road and KeyBank are not involved in winner selection process. The winner will be announced during the AMA Research Challenge final round. Payment of the grand prize will be directly from Laurel Road/KeyBank. See the [AMA Research Challenge Official Rules for terms and conditions](https://www.ama-assn.org/system/files/research-challenge-official-rules.pdf).

[![](https://assets.underline.io/uploads/markdown_image/1/image/611e331d4a0eba0920f240cfc04b4d77.png)](https://www.laurelroad.com/)

#### Code of Conduct
<html>
AMA has a code of conduct for its meetings – we ask all attendees to adhere to this code to participate in this year’s meeting. Please read and <a href="https://www.ama-assn.org/about/code-conduct" target="_blank">learn more here</a> about your conduct requirements.
<br><br>
	</html>
	

#### Semifinals Judges

You need to log in with the email address you registered with. Access credentials have been sent to your email. 

Please be sure to check your spam and other email folders if you do not see an email confirmation right away.

Need help? Contact us at <support@underline.io>.

Please log in to explore this event.

If you have used the AMA Research Challenge platform previously, please click the ‘Log-in’ button below. If you do not remember your password, please click “forgot password” and follow prompts to reset your password. After resetting your password, please refresh your webpage to access the platform. 

Register to explore research in a variety of topics. AMA members get exclusive access to score posters in the semifinals. 

**Please use [this link here](https://www.ama-assn.org/member-benefits/events/ama-research-challenge) to register for free. **

Join the AMA Research Challenge poster symposium and semifinals to explore top research posters and score posters in the semifinals (an AMA member exclusive opportunity).

Neuromodulatory therapies offer promising alternatives for temporal lobe epilepsy (TLE) management, a common and notoriously pharmacoresistant condition. In preclinical TLE models, electrographic-only seizures are defined by abnormal hippocampal electrical activity without overt motor signs. In contrast, behavioral seizures also involve hippocampal epileptiform activity but are additionally accompanied by a progression to motor manifestations such as clonus, falling, or convulsions indicating seizure generalization to extrahippocampal regions. We previously showed that acute, on-demand cerebellar electrical stimulation can robustly suppress hippocampal electrographic seizures. However, this prior study was limited to an acute intervention period and therefore provided insufficient data to assess the impact on the less frequent, overtly behavioral, seizures. Additionally, it remains unknown whether on-demand stimulation retains its efficacy over time, as neural plasticity or network adaptation may impact stimulation responsiveness. Therefore, it is critical to determine whether on-demand cerebellar stimulation remains effective long-term, and whether the intervention also impacts the progression of seizure severity, including the frequency of overtly behavioral seizures events. 

Using the dorsal intrahippocampal kainate mouse model of TLE, we chronically recorded local hippocampal field potentials permitting quick, on-line, detection of spontaneous seizures, and on-demand electrical stimulation was delivered to the midline of the cerebellar cortex (vermis lobules 4/5 and 6). Animals were recorded over a 10-week period composed of a 2-week baseline and an 8-week intervention (or control) period. Animals in the on-demand stimulation condition received intervention selectively at the time of detected seizures while no-stimulation (control) animals received no intervention. 

Preliminary results suggest on-demand cerebellar electrical stimulation remains effective over a chronic period: during the first week of on-demand intervention 51-91% of detected electrographic seizures terminated within 5 seconds of on-demand stimulation; during the final week, 55-98% of detected seizures were terminated (across n=6 animals), indicating continued effectiveness of intervention. Chronic intervention also had positive impacts on behavioral seizures. Animals receiving on-demand intervention to truncate electrographic seizures showed a lower behavioral seizure frequency than no-stimulation animals (on-demand stimulation animals: 1.625 ± 0.56 overtly behavioral seizures per week, n=7 animals, vs no-stimulation: 2.00 ± 0.33 overtly behavioral seizures per week, n=10 animals; p=0.07, Mann-Whitney U Test), despite comparable baseline frequencies (1.85 ± 0.55 vs. 1.88 ± 0.86; p=0.29, Mann-Whitney). 

These encouraging preliminary findings suggest chronic, on-demand cerebellar stimulation not only sustains electrographic seizure suppression, but may also prevent progression toward more severe, overtly behavioral seizure events, highlighting on-demand cerebellar stimulation’s potential as a therapeutic avenue for TLE management.

Investigation of Chronic Cerebellar Electrical Stimulation for Seizure Control In A Mouse Model of Temporal Lobe Epilepsy

Title: Anatomical variation of the middle cerebral artery

Authors: Alexzander Rich1, Ainsley Durning1, Ho Lok1, Cole McNeil1, Dominik Valdez2, Heather F. Smith1,2, Leigha M. Lynch1,2 
1 Arizona College Osteopathic Medicine, Midwestern University, Glendale, AZ, 85308 USA.
2 Department of Anatomy College of Graduate Studies, Midwestern University, Glendale, AZ, 85308 USA. 

Background: Stroke is the fourth leading cause of death in the United States and the third leading cause of death worldwide, according to the CDC and WHO. The middle cerebral artery (MCA) is the most frequently affected vessel in stroke cases. Anatomical variation within the MCA, therefore, can have clinical implications for stroke presentation, diagnosis, and treatment. Thus, quantifying MCA variations within the general population contributes essential information for clinical management of stroke. Current literature suggests that the worldwide frequency of an early branch of the anterior temporal is ~10%, an accessory MCA (hypertrophied artery of Huebner) is 0.1% to 9.1% and a duplicate MCA (additional branch from the internal carotid artery) is 0.3% to 7.1%.

Methods: We collected MCA variation data from the human donors studied within the gross anatomy courses at Midwestern University (Glendale, AZ). The MCA was classified by the presence of either classical MCA presentation, early branching of the anterior temporal branch, accessory MCA, and duplicate MCA​​. We then calculated the frequency of each classification. Both diameter of the MCA within the M1 segment as well as the length of the M1 segment were also measured. Left and right middle cerebral arteries were compared between sexes (N=28 male, N=20 female). 

Results: The frequencies, when both sexes are included, of left sided MCA variation was 85.4% classical, 8.3% early branching of anterior temporal, 0% accessory and 0% duplicate. The frequencies of right sided MCA variation was 83.3% classical, 4.2% early branching of anterior temporal, 6.3% accessory and 0% duplicate. We found that the mean diameter on the right MCA significantly differed between sexes, with a male mean diameter of 3.04 mm and female mean diameter 3.38 mm (p=0.0275). Pearson correlation coefficients were calculated to assess the relationship between left and right MCA diameter and length. We found a moderate correlation between left and right MCA diameter (r=0.549, p=0.0036) and left and right MCA length (r=0.459, p=0.018). There is no correlation between MCA length and diameter (left p=0.060, right p=0.213).

Conclusion: These findings suggest that further understanding of MCA variations during stroke diagnosis and clinical management is necessary to decrease the time of MCA occlusion to improve patient outcome. Additional data and assessment of how these branching patterns alter the cortical regions affected by stroke will also aid in improving patient treatment.

Middle Cerebral Artery: Variable Presentation Within a Donor Population

Background: Videolaryngostroboscopy (VLS) is the gold standard for assessing laryngeal function in voice disorders. Unlike traditional laryngoscopy, it uses stroboscopic light to create slow-motion views of vocal fold (VF) vibrations, capturing synchronized high-quality video and audio of voice production. Interpreting VLS examinations and diagnosing VF pathologies often relies on subjective assessments and may not fully capture their complexity. Recent advancements in Artificial Intelligence (AI) have opened new avenues for enhancing diagnostic accuracy by integrating video, audio, and text data, more closely approximating clinician evaluation. This study aimed to develop and validate a multimodal deep-learning classifier trained on stroboscopic images, voice, and clinicodemographic data, leveraging each modality’s strengths to differentiate among three distinct VF states: healthy (HVF), unilateral paralysis (UVFP), and VF lesions, including benign and malignant pathologies. 

Methods: VLS recordings were collected from adult patients seen at the Sean Parker Institute for Voice between January 2018 and December 2022. Patients with UVFP (n = 54), VF lesions (n = 42), and HVF (n = 41) were retrospectively identified. Image frames and voice samples were extracted from stroboscopic videos, including sustained /i/ phonation, pitch glide, and /i/ sniff task. Clinicodemographic variables were collected from the electronic health record. Visual features were extracted using vision transformer technology, DINOv2, and acoustic features were extracted using Librosa. All three modalities were combined using an Artificial Neural Network. Unimodality models using image or voice data were trained for comparison. Patient-level data was independently divided into training (80%) and testing (20%). Accuracy, F1-scores, and AUROC were used to validate models. 

Results: When evaluated on a hold-out test set, the multimodal classifier demonstrated an overall accuracy of 76.92%, with F1-scores of 0.67 for HVF, 0.88 for UFVP, and 0.71 for VF lesions. The corresponding macro-average AUROC was 0.8229. The image classifier achieved an overall accuracy of 61.54%, with F1-scores of 0.50 for HVF, 0.92 for UFVP, and 0.62 for VF lesions, and a macro-average AUROC of 0.7675. The audio classifier achieved an overall accuracy of 65.38%, with F1-scores of 0.62 for HVF, 0.54 for UFVP, and 0.80 for VF lesions, and a macro-average AUROC of 0.7581.

Conclusion: This novel multimodal AI classifier strength in performance compared to image-only model and audio-only models demonstrates the potential for combining stroboscopic frames, voice and text data for classification of VF pathology. Future models should validate these findings on larger datasets.

A Multimodal Approach for Deep-Learning Classification of Vocal Fold Pathologies in Stroboscopy 


Background
Early-onset psychosis (EOP), defined by the presence of psychotic symptoms before the age of 18, is associated with poorer clinical outcomes (cognitive impairment, hospitalizations, psychosocial function) as compared to individuals with adult-onset psychosis. Previous research found that EOP patients frequently report distressing pain and somatosensory symptoms, and that the severity of these symptoms is associated with the degree of psychopathology. This study expands upon those findings by exploring the prevalence and timing of pain and sensory symptoms in EOP patients.

Methods
A retrospective review of 77 patients (45 male, 32 female) with a DSM-5 diagnosis of psychotic disorder before age 18 and no comorbid autism spectrum disorder was conducted. Patients were grouped by age at psychosis onset: <8 years, 8–13 years, and 14–18 years. All available clinical notes in the electronic health record were reviewed. Data on frequent and severe pain symptoms (musculoskeletal, headache, chest, gastrointestinal) without an identifiable cause and aversive sensory symptoms (auditory/visual, tactile, taste/smell) based on the Short Sensory Profile were abstracted. Patients were categorized into four groups: pain only, sensory sensitivities only, both symptoms, and neither. The timing of pain and sensory symptoms relative to psychosis onset (before/same-time or after) was also recorded. All cases were reviewed and confirmed by an expert clinician.

Results
Of the 77 patients with EOP, 95% (n=73) reported pain, sensory symptoms or both. The majority experienced pain and sensory symptoms; onset <8 (n=21): 67%, onset 8-13 (n=41): 51%, and aged 14-18 (n=15): 80%.

Regarding symptom timing, among the 66 patients reporting pain, 73% (n=48) experienced symptoms before or at the time psychosis was first identified. Of the 40 patients with sensory symptoms, 98% (n=39) experienced symptoms by the time psychosis was recognized. Headache, frequent or severe enough to be reported to a clinician, was the most common pain symptom (76% with pain), and aversive auditory/visual sensitivities were the most common sensory symptom (62% with sensory symptoms).

Conclusion
This study expands upon prior findings by providing evidence that these somatosensory symptoms are typically present by the time the psychosis is recognized. The high prevalence of pain or sensory symptoms causing sufficient distress to prompt medical attention suggests that targeted treatment of these symptoms should be considered in the care of EOP patients.

Pain and Sensory Symptoms as Early Clinical Features of Early Onset Psychosis

Background

Salmonella enterica is an uncommon cause of urinary tract infections (UTIs), particularly in the absence of gastrointestinal symptoms. When present in urine cultures, non-typhoidal Salmonella should prompt investigation for extra-urinary reservoirs. The gallbladder can act as a chronic reservoir in patients with cholelithiasis, even without active cholecystitis, and may serve as a source of intermittent bacterial seeding. Recognition of this rare presentation is critical, as it poses diagnostic and management challenges and may require public health notification.

Case Presentation

An 86-year-old male with a history of bilateral nephrolithiasis, suprapubic catheterization, and cholelithiasis presented with recurrent UTIs over a three-month period. In February 2025, urine cultures repeatedly grew non-typhoidal Salmonella enterica (>100,000 CFU/mL), confirmed via MALDI-TOF. Despite appropriate IV antibiotic therapy with ceftriaxone, the patient’s symptoms persisted. Stool cultures were also positive for Salmonella enterica, supporting chronic gastrointestinal carriage despite the absence of diarrhea, recent travel, or sick contacts.

CT imaging revealed cholelithiasis with biliary sludge but no acute cholecystitis. Importantly, imaging also ruled out a colovesical fistula, eliminating a direct anatomical connection between the gastrointestinal and urinary tracts. With negative blood cultures and no signs of systemic bacteremia, these findings supported an ascending route of urinary tract infection. His clinical course was marked by persistent leukocytosis, sterile pyuria, and suprapubic discomfort.

Infectious disease consultation identified chronic gallbladder carriage as the most likely source of recurrent urinary Salmonella. Given the patient’s elevated surgical risk, cholecystectomy was deferred, and he was managed with long-term suppressive antibiotics. The case was reported to hospital infection control and the state health department for public health surveillance and serotyping.

Discussion

This case illustrates a rare but clinically significant presentation of recurrent UTIs due to chronic gallbladder Salmonella enterica carriage. The diagnosis was supported by microbiologic confirmation from urine and stool cultures and ruled-out anatomic causes of direct contamination. In patients with biliary pathology and persistent infection with uncommon urinary pathogens, hepatobiliary colonization should be considered. Early infectious disease involvement, careful imaging, and public health coordination were essential for diagnosis and management. Recognition of this atypical infection pathway emphasizes the importance of pathogen-specific investigation in persistent or recurrent UTIs.

Recurrent Urinary Tract Infections from Chronic Gallbladder Salmonella enterica Carriage

Background:
FLT3-ITD mutation occurs in 25–30% of acute myeloid leukemia (AML) and is associated with a poor prognosis. FLT3 inhibitors (FLT3i), such as gilteritinib and quizartinib, improve initial response rates, but relapse is common. While on-target resistance via FLT3-TKD mutations is established, emerging evidence suggests that “collateral resistance,” characterized by the outgrowth of non-FLT3 subclones, contributes significantly to therapeutic escape. This review synthesizes current data on clonal evolution under FLT3i pressure, focusing on non-FLT3 subclonal emergence.

Methods:
We conducted a comprehensive review of published studies done up to June 2025, in PubMed, Web of Science, and Scopus. Studies were included if they involved FLT3-mutated AML patients treated with FLT3 inhibitors, reported paired genomic data at diagnosis and relapse. Data on FLT3i type, resistance mechanisms, mutation spectrum, and variant allele frequency (VAF) changes were extracted from clinical trials, retrospective cohorts, and preclinical models.

Results:
In a cohort of 41 relapsed patients treated with gilteritinib, 33% developed FLT3-TKD mutations (D835, F691L), 36.6% acquired NRAS/KRAS mutations, and 5% developed BCR-ABL1 fusions. Another study found off-target RAS/MAPK pathway mutations in 30–37% of relapses, frequently co-occurring with ASXL1, RUNX1, and PTPN11 mutations. Resistance to quizartinib was driven by FLT3-TKD mutations in 33–50% of cases, yet up to 50% of relapses lacked FLT3 mutations, suggesting alternative mechanisms such as microenvironmental signaling and off-target mutations. Three reported cases of BCR-ABL1 fusion following FLT3i failure showed partial responses to combined FLT3 and ABL1 inhibition. Single-cell DNA sequencing revealed that non-FLT3 subclones, undetectable at baseline by bulk sequencing, expanded under FLT3 inhibitor pressure, demonstrating polyclonal resistance and clonal selection as key drivers of relapse. These studies consistently demonstrate that non-FLT3 resistance, particularly involving the RAS/MAPK axis, is frequent and clinically relevant following targeted therapy.

Conclusion:
Collateral resistance to FLT3 inhibitors in AML is frequently mediated by the selection of pre-existing, low-frequency non-FLT3 mutant subclones, particularly those involving the RAS/MAPK pathway. These findings emphasize the importance of comprehensive genomic profiling for diagnosis and relapse. Future therapeutic strategies should prioritize rational drug combinations targeting FLT3 and non-FLT3 pathways to overcome resistance and improve patient outcomes.


REFERENCES
1.Smith, C. C., Levis, M. J., Perl, A. E., et al. (2022). Molecular profile of FLT3-mutated relapsed/refractory patients with AML in the phase 3 ADMIRAL study of gilteritinib. Blood Advances, 6(7), 2144–2155. https://doi.org/10.1182/bloodadvances.2021006489 (Leading Journal)
2. Smith, C. C., Lin, K., Stecula, A., Sali, A., & Shah, N. P. (2015). FLT3 D835 mutations confer differential resistance to type II FLT3 inhibitors. Leukemia, 29(12), 2390–2392. https://doi.org/10.1038/leu.2015.165 (Leading Journal)
3. Tarver, T. C., Hill, J. E., Rahmat, L., et al. (2020). Gilteritinib is a clinically active FLT3 inhibitor with broad activity against FLT3 kinase domain mutations. Blood Advances, 4(3), 514–524. https://doi.org/10.1182/bloodadvances.2019000919 (Leading Journal)



Collateral Resistance in FLT3-Mutated AML: Unraveling Non-FLT3 Subclonal Emergence and Therapeutic Escape Under FLT3 Inhibitor Pressure - A Comprehensive Literature Review

Background: The compression of pre-clinical medical education from 24 to 18 months has pushed students to rely on third-party resources (TPR) for board preparation. While TPRs are effective, they are expensive and numerous. Choosing a TPR can feel overwhelming for students still adjusting to academic demands. Despite widespread use, limited data guides TPR selection. Prior studies have described usage patterns, but none have examined whether learning styles predict resource preference, and no tools currently exist to guide platform choice. We aimed to characterize current TPR use and determine whether learning style, as measured by the Barsch Learning Style Inventory (BLSI), could predict resource preference.

Methods: A cross-sectional REDCap survey was distributed to MD students (Classes of 2025–2028) at a single U.S. institution. Students ranked their top three resources for learning and reviewing material, reported reasons for using TPRs, and completed the validated BLSI. Based on their highest BLSI scoring domain, students were categorized as visual, auditory, or kinesthetic learners. Trends in resource utilization were assessed across the cohort and within each learning style using Fisher’s Exact Test with Benjamini-Hochberg correction for pairwise comparisons.

Results: Of 149 responses, 104 were included in the analysis; 45 incomplete responses were excluded. The cohort included 67 visual, 24 kinesthetic, and 13 auditory learners. Over 80% of students trialed TPRs including Sketchy, First Aid (FA), Boards and Beyond (B&B), UWorld, Anki, and Bootcamp, while only 72% trialed Institution-Provided Resources (IR). B&B (71%), Bootcamp (45%), IR (30%), and Sketchy (38%) were the most frequently selected learning resources, based on the proportion of students who included them in their top three rankings. Visual learners showed a significant preference for B&B over other learning resources (p<0.01). Auditory and kinesthetic learners also favored B&B and Bootcamp, though differences were not statistically significant. Analysis of review resources revealed no significant differences across learning styles, with UWorld, Anki, and FA consistently ranking as the top three choices. Overall, 95.5% of participants indicated they would choose a TPR over IRs if limited to one for Step 1 preparation.

Conclusion: Learning style may influence preferred TPRs, with visual learners showing a significant inclination toward B&B. While B&B and Bootcamp were broadly favored, students often trial multiple TPRs at personal expense before settling on one. A predictive tool using BLSI may help streamline TPR selection. Institutions may leverage these findings to support student decision-making and subsidize high-value resources, reducing financial waste.

Predicting Third-Party Resource Preferences Using the Barsch Learning Style Inventory

Introduction: Opill’s 2023 FDA approval for over-the-counter (OTC) sale could expand contraceptive access in the United States without age restrictions. Since approval does not ensure widespread access, we sought to evaluate Opill’s availability in New York State. 

Methods: We randomly sampled 379 of 7143 New York pharmacies from the National Provider Identification directory to estimate the proportion stocking Opill with 95% confidence. We classified pharmacies by type (independent, chain, or chain-in-retailer), location (metro or non-metro), and county-level 2024 presidential election results. Posing as 16-year-old secret shoppers, we collected on in-store availability, ordering availability, cost, and age restrictions. We performed descriptive statistical analyses and Chi-squared testing, setting significance at 5% using Stata SE Version 16.0. 

Results: In June-July 2025, 214/379 (56.5%) New York pharmacies had Opill in-store. Of those that did not, 76/165 (46.1%) could order it with a one day median delivery time. The average one month cost was $25.29. Chain pharmacies were more likely to stock Opill (76.5% vs. 52.8% chain-in-retailer vs. 46.05% independent; p<0.001). Availability did not differ by urban status or county political affiliation. Additionally, 21 (7.3%) pharmacists imposed age restrictions, 36 (9.5%) confused Opill with emergency contraception, 16 (4.2%) never heard of Opill, and 9 (2.4%) ceased stocking given low sales. 

Conclusions: Current OTC availability of Opill appears limited to approximately half of New York State pharmacies. Knowledge gaps among pharmacy staff should be addressed to maximize Opill’s public health impact.

Over-the-Counter, Out of Reach: A Secret Shopper Study of Opill Availability in New York Pharmacies


Background
Alzheimer's Disease (AD), a neurodegenerative condition causing progressive cognitive and behavioral decline, is the most prevalent dementia, accounting for roughly 70% of all cases. Its pathology is complex, influenced by genetic, lifestyle, and environmental factors. Affecting nearly seven million Americans aged 65 and older and ranking as the fifth-leading cause of death, AD mortality has risen, particularly in rural areas due to limited healthcare access. This barrier often leads to misdiagnosis or delayed AD diagnosis, worsening patient outcomes. This review examines three major AD healthcare access barriers for the rural geriatric population and proposes early preventive interventions to enhance patient quality of life.

Methods
A literature search was performed across several academic databases, including PubMed and ScienceDirect. Relevant studies included peer-reviewed articles published between the years 2020 to 2025. Key search terms included “Alzheimer’s”, “rural”, “barriers to care”, and “prevention”.

Results
Rural living presents significant healthcare access challenges for AD patients requiring specialized treatment. Primary obstacles include limited healthcare infrastructure, with many towns lacking hospitals or solely having clinics with restricted schedules. This limitation, therefore, forces patients to travel significant distances to receive medical care. Building on that point, transportation further aggravates barrier issues. This is because older adults often lack vehicles or rely on caregivers with conflicting schedules due to medical or socioeconomic challenges. Lastly, insufficient digital access remains a major barrier. Despite telemedicine expansion post-COVID-19, its utilization by the geriatric population is hampered by limited digital literacy, absent home internet, and inadequate navigation support in many rural regions.
To mitigate these disparities, several actions are crucial for AD preventive care. Firstly, promoting cognitive stimulation through senior community centers can reduce AD risk by combating isolation. Secondly, improving access to affordable, nutritious local foods is vital for low-income rural individuals, impacting cognitive health. Thirdly, improving physical exercise via expanded recreation centers or sidewalks fosters safe activity. Finally, enhancing AD education for patients and caregivers is paramount, facilitating earlier symptom recognition and diagnosis during prodromal stages, thereby improving quality of life through timely intervention.

Conclusion
AD is a complex, incurable condition, necessitating more research to optimize treatment, especially for rural populations. Key barriers, including inadequate facilities, transportation, and digital illiteracy, can be addressed through initiatives like building more rural hospitals, expanding bus routes, and offering local technology education. Through expanded research, education, advocacy, and awareness, AD prevention, early diagnosis, intervention, and management will undoubtedly improve across rural communities in the United States and beyond.


Bridging the Divide: Addressing Alzheimer's Disease Healthcare Barriers in Rural America

Abstract Title
Beyond Screening: A Cross-National Study of Gender Equity, Health Spending, and Mammography Infrastructure

Baylor Akhavan-Mora B.S.1, Sherli Koshy-Chenthittayil Ph.D.2, Amy E.L. Stone Ph.D.1

1Department of Medical Education, Kirk Kerkorian School of Medicine at the University of Nevada, Las Vegas
2Office of Institutional Effectiveness, Touro University Nevada

Background
Patient outcomes in oncology improve with early detection, and mammography units play a critical role in identifying neoplasms before metastasis. However, access may be limited in countries with lower development or greater gender inequality. Is the national density of mammography units associated with the Gender Development Index (GDI) across WHO member states, and how does this relate to health expenditures?

Methods
Publicly available data from the WHO Global Health Observatory and UNDP were used to explore the relationship between mammography unit density, GDI, and health expenditures as a percentage of GDP. Using R, Pearson correlation and multiple linear regression analyses were performed.

Results
Data from 129 WHO member states (2014–2021) showed a moderate positive correlation between GDI and mammography unit density (r = 0.34, 95% CI [0.18, 0.49], p < 0.0001), suggesting that countries with greater gender parity tend to invest more in women’s preventative healthcare. A stronger correlation was found between health expenditures and mammography density (r = 0.43, 95% CI [0.28, 0.56], p < 0.0000004). However, multiple linear regression showed that GDI and health expenditures explained only 25% of the variance in mammography unit density, indicating additional influencing factors.

Conclusion
While GDI and health expenditures are significantly associated with mammography access, over 75% of the variance remains unexplained. Future research should investigate other sociopolitical, geographic, and health system factors to better understand disparities in screening infrastructure.


Premium content

Next from AMA Research Challenge 2025

Investigation of Chronic Cerebellar Electrical Stimulation for Seizure Control In A Mouse Model of Temporal Lobe Epilepsy

Similar lecture

Predicate Lineages and Medical Device Recalls in the 510(k) Pathway