Daily Archives: December 25, 2024
VisionFM: A Generalist AI Surpasses Single-Modality Models in Ophthalmic Diagnostics
25 Dec, 2024 | 13:41h | UTCBackground: Ophthalmic AI models typically address single diseases or modalities. Their limited generalizability restricts broad clinical application. This study introduces VisionFM, a novel foundation model trained on 3.4 million images from over 500,000 individuals. It covers eight distinct ophthalmic imaging modalities (e.g., fundus photography, OCT, slit-lamp, ultrasound, MRI) and encompasses multiple diseases. Compared with prior single-task or single-modality approaches, VisionFM’s architecture and large-scale pretraining enable diverse tasks such as disease screening, lesion segmentation, prognosis, and prediction of systemic markers.
Objective: To develop and validate a generalist ophthalmic AI framework that can handle multiple imaging modalities, recognize multiple diseases, and adapt to new clinical tasks through efficient fine-tuning, potentially easing the global burden of vision impairment.
Methods: VisionFM employs individual Vision Transformer–based encoders for each of the eight imaging modalities, pretrained with self-supervised learning (iBOT) focused on masked image modeling. After pretraining, various task-specific decoders were fine-tuned for classification, segmentation, and prediction tasks. The model was evaluated on 53 public and 12 private datasets, covering eight disease categories (e.g., diabetic retinopathy, glaucoma, cataract), five imaging modalities (fundus photographs, OCT, etc.), plus additional tasks (e.g., MRI-based orbital tumor segmentation). Performance metrics included AUROCs, Dice similarity coefficients, F1 scores, and comparisons with ophthalmologists of varying clinical experience.
Results: VisionFM achieved an average AUROC of 0.950 (95% CI, 0.941–0.959) across eight disease categories in internal validation. External validation showed AUROCs of 0.945 (95% CI, 0.934–0.956) for diabetic retinopathy and 0.974 (95% CI, 0.966–0.983) for AMD, surpassing baseline deep learning approaches. In a 12-disease classification test involving 38 ophthalmologists, VisionFM’s accuracy matched intermediate-level specialists. It successfully handled modality shifts (e.g., grading diabetic retinopathy on previously unseen OCTA), with an AUROC of 0.935 (95% CI, 0.902–0.964). VisionFM also predicted glaucoma progression (F1, 72.3%; 95% CI, 55.0–86.3) and flagged possible intracranial tumors (AUROC, 0.986; 95% CI, 0.960–1.00) from fundus images.
Conclusions: VisionFM offers a versatile, scalable platform for comprehensive ophthalmic tasks. Through self-supervised learning and efficient fine-tuning, it extends specialist-level performance to multiple clinical scenarios and imaging modalities. The study demonstrates that large-scale, multimodal pretraining can enable robust generalization to unseen data, potentially reducing data annotation burdens and accelerating AI adoption worldwide.
Implications for Practice: VisionFM may help address global shortages of qualified ophthalmologists and expand care in low-resource settings, though clinical decision-making still requires appropriate human oversight. Further multicenter studies are needed before widespread implementation, especially for higher-risk use cases such as tumor detection.
Study Strengths and Limitations: Strengths include its unique multimodal design, large-scale pretraining, and extensive external validation. Limitations involve demographic bias toward Chinese datasets, the need for larger cohorts in certain applications (e.g., intracranial tumor detection), and the challenges of matching real-world clinical complexity when only image-based data are used.
Future Research: Further validation in diverse populations, integration of new imaging modalities (e.g., widefield imaging, ultrasound variants), and expansion to additional diseases are planned. Hybridization with large language models could facilitate automatic generation of clinical reports.
Reference: Qiu J, Wu J, Wei H, et al. Development and Validation of a Multimodal Multitask Vision Foundation Model for Generalist Ophthalmic Artificial Intelligence. NEJM AI 2024;1(12). DOI: http://doi.org/10.1056/AIoa2300221
RCT: Avoiding Prophylactic Drain Increases Postoperative Invasive Procedures After Gastrectomy
25 Dec, 2024 | 12:47h | UTCBackground: Prophylactic abdominal drainage following gastrectomy for gastric cancer has been debated for decades. While some Enhanced Recovery After Surgery (ERAS) guidelines discourage routine drains, many surgeons still advocate their use to detect and manage intra-abdominal collections before they become severe. Previous trials were small and underpowered, thus failing to provide robust evidence regarding the real need for prophylactic drains.
Objective: To determine whether omitting a prophylactic drain in gastric cancer surgery leads to a higher likelihood of postoperative invasive procedures (reoperation or percutaneous drainage) within 30 days.
Methods: In this multicenter randomized clinical trial, 404 patients from 11 Italian centers were randomly assigned to either prophylactic drain placement or no drain at the end of subtotal or total gastrectomy. Both academic and community hospitals participated. The primary composite outcome was the rate of reoperation or percutaneous drainage within 30 postoperative days, analyzed via a modified intention-to-treat approach. Secondary endpoints included overall morbidity, anastomotic leaks, length of hospital stay, and 90-day mortality. A parallel invited commentary addressed methodological and clinical perspectives.
Results: Among the 390 patients who underwent resection, 196 had a prophylactic drain and 194 did not. By postoperative day 30, 7.7% of patients in the drain group required reoperation or percutaneous drainage, compared with 15% in the no-drain group. This statistically significant difference was driven by a higher reoperation rate in patients without drains. Both groups had similar anastomotic leak rates (approximately 4% overall). However, patients without prophylactic drains had a higher in-hospital mortality (4.6% vs 0.5%) and were more likely to require escalation of care. There were few drain-related complications, indicating a low risk associated with drain placement. Length of hospital stay and readmission rates were comparable between groups.
Conclusions: Omitting prophylactic drains in gastrectomy was associated with an increased need for postoperative invasive interventions, particularly reoperations. While prior guidelines have recommended against routine drain placement, these findings challenge that stance for total and even subtotal gastrectomies. Surgeons may wish to revisit existing protocols, especially in facilities with fewer resources or lower patient volumes, given the potential reduction in reoperation risk associated with prophylactic drainage.
Implications for Practice: Clinicians should carefully balance possible benefits (earlier detection of fluid collections and reduced reoperations) against potential drawbacks of drain usage. Routine placement may be reconsidered, at least in higher-risk cases or in institutions less equipped for complex salvage procedures.
Study Strengths and Limitations: Key strengths include its robust sample size and standardized criteria for complications. Limitations involve the unblinded nature of postoperative management and the lack of drain fluid amylase measurements to guide removal protocols. Additionally, differentiating total from subtotal gastrectomies might refine selection criteria for prophylactic drainage.
Future Research: Further studies could focus on stratified risk profiles for total vs subtotal gastrectomy and on biomarkers in drain fluid to identify subgroups most likely to benefit from prophylactic drainage.
Meta-analysis: Incidence Rate Difference of Adverse Events from Canabinoids in Middle-Aged and Older Adults
25 Dec, 2024 | 12:19h | UTCBackground: Growing evidence suggests that cannabinoid-based medicines (CBMs) are increasingly prescribed to individuals aged 50 years and above for various clinical conditions. While these agents may offer therapeutic benefits, questions remain about the incidence of adverse events (AEs), particularly in older adults with multiple comorbidities. This systematic review and meta-analysis aims to quantify the incidence rate difference (IRD) of AEs and determine whether weekly doses of delta-9-tetrahydrocannabinol (THC) and cannabidiol (CBD) are associated with any dose-dependent increase in risk.
Objective: To evaluate whether adults aged ≥50 years exposed to CBMs, including THC-alone formulations and THC combined with CBD, experience a higher incidence of AEs than controls, and to assess how variations in weekly THC and CBD doses might affect AE rates.
Methods: Researchers searched MEDLINE, PubMed, EMBASE, CINAHL, PsychInfo, Cochrane Library, and ClinicalTrials.gov from January 1, 1990, to June 12, 2023. Randomized clinical trials involving middle-aged and older adults (mean age ≥50 years) using medicinal CBMs for all indications were included. Data on common and serious AEs, withdrawals, and deaths were extracted and pooled using a random-effects model. Further meta-regression analyses examined THC and CBD weekly doses as predictors of AEs in THC-only and THC:CBD trials.
Results: Fifty-eight randomized clinical trials (n=6611) met inclusion criteria, with 3450 participants receiving CBMs. Compared to controls, individuals on THC-alone experienced significantly higher incidence of dizziness, somnolence, impaired coordination, dissociative symptoms, and dry mouth, often in a dose-dependent manner. Similarly, THC:CBD combinations increased nausea, vomiting, fatigue, dizziness, and disorientation. The incidence of serious AEs, withdrawals, or mortality did not differ significantly between CBM and control groups, although neurological or psychiatric side effects were more pronounced with higher THC doses.
Conclusions: THC-containing CBMs can provoke dose-related gastrointestinal, neurological, and psychiatric adverse events, posing additional risks in older adults susceptible to falls and cognitive disturbances. However, the meta-analysis found no significant increases in serious AEs or deaths. Clinicians should weigh potential benefits against the likelihood of common side effects, especially when prescribing higher THC doses or combining cannabinoids with other medications frequently used by older patients.
Implications for Practice:
- Physicians should exercise caution when initiating or escalating THC-based therapies in middle-aged and older adults, monitoring for neurological or psychiatric AEs.
- Using lower THC doses, titrating gradually, and adding CBD may mitigate some side effects.
- Though severe AEs are uncommon, vigilance is warranted in individuals with complex medication regimens.
Study Strengths and Limitations:
- Strength: This review merges diverse clinical conditions and provides a comprehensive assessment of THC vs. THC:CBD. Its large pooled population allows for more precise IRD estimates.
- Limitation: Short treatment durations in many trials limit understanding of long-term toxicity, and some studies lacked rigorous reporting of randomization and outcome measures, potentially introducing bias.
Future Research:
- Longer-duration trials focused on older populations are needed to clarify chronic safety profiles.
- Studies exploring drug-drug interactions between CBMs and medications commonly prescribed to older adults will further elucidate real-world tolerability.
Reference: Velayudhan L, Pisani S, Dugonjic M, McGoohan K, Bhattacharyya S. Adverse events caused by cannabinoids in middle aged and older adults for all indications: a meta-analysis of incidence rate difference. Age and Ageing. 2024;53(11):afae261. DOI: https://doi.org/10.1093/ageing/afae261
Bayesian Network Meta-Analysis: Chlorpromazine IV/IM Emerges as a Top Choice for Acute Migraine Relief in the ED
25 Dec, 2024 | 11:18h | UTCBackground: Acute migraine is a prevalent cause of emergency department (ED) visits, necessitating prompt pain control. Although numerous drugs are available, there is debate about the most effective and safest options. Traditional pairwise meta-analyses fail to capture all treatment comparisons in a single framework, making network meta-analyses, particularly Bayesian, an appealing approach to inform clinical decision-making.
Objective: This systematic review and Bayesian network meta-analysis aimed to compare multiple pharmacologic therapies—single agents or combinations—for acute migraine relief in adults presenting to the ED. The goal was to identify those most likely to achieve adequate pain relief, reduce rescue medication use, and minimize significant adverse reactions.
Methods: The authors searched MEDLINE, Embase, and Web of Science from inception to February 9, 2024, for randomized controlled trials comparing any pharmacologic therapy to another or to placebo in ED patients with migraine. Four primary outcomes were analyzed: (1) adequate pain relief at two hours, (2) change in pain intensity at one hour, (3) need for rescue drug at two hours, and (4) significant adverse reaction (eg, sedation, akathisia, hypotension).
Results: Twenty-four to twenty-seven trials contributed to each outcome network. Chlorpromazine IV/IM was ranked highest for adequate pain relief (SUCRA=87.3%) and also significantly reduced the need for rescue medication (SUCRA=93.2%). Ibuprofen IV and valproate IV emerged among the least effective for pain relief, while dexamethasone IV was the most probable to cause fewer serious adverse reactions (SUCRA=79.5%). However, most comparisons were of low or very low certainty, limiting the strength of the findings.
Conclusions: Chlorpromazine IV/IM appears among the most effective single agents for acute migraine in the ED, although it may carry higher risks of sedation or hypotension. Certain analgesics (eg, ibuprofen IV, valproate IV, and possibly ketorolac IV/IM) demonstrated lower efficacy. Due to variability in trial size, dosing, and participant characteristics, the overall certainty of evidence remains limited.
Implications for Practice: Clinicians may consider parenteral chlorpromazine for rapid migraine relief, balancing its adverse event profile with potential efficacy. Dexamethasone’s lower probability of serious side effects could make it a complementary option. The findings highlight the need for individualized treatment, taking into account patient comorbidities and preferences.
Study Strengths and Limitations: This network meta-analysis offers a broad comparative perspective on diverse pharmacologic interventions for ED-based migraine management. Nonetheless, there is notable heterogeneity in study methodologies, small sample sizes, and sparse direct comparisons for many interventions, all of which reduce certainty in the estimates.
Future Research: Larger, more standardized trials are needed to confirm these results and directly compare drugs like chlorpromazine, prochlorperazine, and metoclopramide-NSAID combinations. Rigorous safety reporting is crucial to clarify adverse reaction risks for various agents, especially those with less available evidence.
Reference: deSouza IS, Anthony N, Thode H Jr, et al. Effectiveness and Safety of Pharmacologic Therapies for Migraine in the Emergency Department: A Systematic Review and Bayesian Network Meta-analysis. Annals of Emergency Medicine. DOI: http://doi.org/10.1016/j.annemergmed.2024.11.004
AGA Clinical Practice Update on Screening and Surveillance in High-Risk US Populations for Gastric Cancer: Expert Review
25 Dec, 2024 | 11:02h | UTCIntroduction:
This American Gastroenterological Association (AGA) Clinical Practice Update provides guidance on primary and secondary prevention strategies for gastric cancer (GC) among high-risk groups in the United States. GC disproportionately affects racial and ethnic minorities, certain first-generation immigrants from countries with elevated GC incidence, and individuals with specific hereditary syndromes or family histories of GC. Given ongoing disparities in diagnosis and outcomes, this document outlines best practices for recognizing at-risk individuals, performing high-quality endoscopic screening, and establishing surveillance protocols for gastric precancerous conditions.
Key Recommendations:
- Identify High-Risk Groups: Consider screening among first-generation immigrants from high-incidence regions, people with a family history of GC in a first-degree relative, individuals with hereditary gastrointestinal syndromes, and patients with multiple risk factors (eg, chronic Helicobacter pylori infection, smoking, diets high in salt and processed meats).
- Preferred Screening Modality: Upper endoscopy is considered the best method for detecting precancerous lesions (atrophic gastritis and intestinal metaplasia) and early malignancies. It allows direct visualization of the gastric mucosa, systematic biopsy, and accurate histologic staging.
- High-Quality Endoscopic Examination: Essential elements include high-definition endoscopes, optimal mucosal cleansing and insufflation, adequate inspection time, systematic photodocumentation, and biopsy protocols (such as the updated Sydney System) to detect and characterize precancerous changes or early cancer.
- H. pylori Eradication: Opportunistic screening for H. pylori and its eradication are key adjunctive measures in preventing GC development. Family-based testing—screening adult household members of H. pylori–positive individuals—may further reduce reinfection rates and disease progression.
- Systematic Biopsy Protocols: When atrophic gastritis or intestinal metaplasia is suspected, obtain at least five biopsies (antrum/incisura and corpus in separate containers). Any suspicious lesion should be sampled independently.
- Recognition of Metaplasia and Dysplasia: Endoscopists should be trained to accurately identify visual patterns associated with gastric intestinal metaplasia (GIM) and dysplasia. Artificial intelligence may hold promise, but current data are insufficient to recommend routine use.
- Risk Stratification and Surveillance Intervals: Patients with confirmed GIM or dysplasia, especially those with severe or extensive metaplasia, may require follow-up endoscopy every three years. Individuals with multiple risk factors or severe metaplastic changes could benefit from shorter intervals.
- Management of Dysplasia and Early GC: All dysplasia should be reviewed by an expert gastrointestinal pathologist. Visible high-grade dysplasia or early GC lesions generally warrant endoscopic submucosal dissection (ESD) at specialized centers to achieve en bloc, R0 resection and enable accurate pathology.
- Post-Resection Surveillance: Individuals with successfully resected dysplasia or early cancer need ongoing endoscopic surveillance to detect metachronous lesions. Surveillance intervals vary depending on pathology results and patient-level factors.
- De-Escalation of Screening: Discontinue screening or surveillance when the patient is no longer fit for potential endoscopic or surgical treatment.
- Equity and Sustainability: To reduce GC mortality, it is crucial to address modifiable risk factors, enhance patient access to endoscopy and skilled practitioners, and integrate research advances, especially in noninvasive biomarker development and improved endoscopic technologies.
Conclusion:
An effective US-based GC screening and surveillance program requires robust preprocedural identification of high-risk individuals, intraprocedural adherence to quality endoscopy standards, and consistent postprocedural follow-up to ensure equitable access to treatment. By refining these clinical practices and prioritizing research, meaningful reductions in GC incidence and mortality can be achieved, ultimately improving patient outcomes and addressing healthcare disparities.
Reference:
Shah SC, Wang AY, Wallace MB, Hwang JH. AGA Clinical Practice Update on Screening and Surveillance in Individuals at Increased Risk for Gastric Cancer in the United States: Expert Review. Gastroenterology. Published online December 23, 2024.
https://doi.org/10.1053/j.gastro.2024.11.001