Author: admin

Posted on

STAT+: Enliven Therapeutics’ leukemia drug shows promise in new study

A targeted drug from Enliven Therapeutics induced molecular responses in nearly half of patients with advanced leukemia, including higher response rates in patients treated at an earlier stage of their disease.

The updated early-stage study results reported Thursday for the Enliven drug, ELVN-001, compare favorably to a current blockbuster medicine sold by Novartis and an upstart experimental drug recently bought by Merck.

At 24 weeks, an 80 mg, once-daily dose of ELVN-001 achieved a major molecular response in 48% of patients with chronic myeloid leukemia, or CML, a slow-growing cancer that starts in myeloid cells.

Continue to STAT+ to read the full story…

Posted on

STAT+: ‘Synthetic lethality’ could trigger another round of biotech M&A

This story first appeared in Adam’s Biotech Scorecard, a subscriber-only newsletter. STAT+ subscribers can sign up here to get it delivered to their inbox.

Never before have I covered so much positive news about pancreatic cancer in such a short period of time. What happens next? Could Revolution Medicines buy Tango Therapeutics? Or, perhaps Bristol Myers Squibb goes all out and acquires Revolution Medicines?

To be clear, neither of these deals has been announced, or even rumored. I’m just playing the biotech M&A speculation game. But a strong case for something to happen can be made in the wake of Monday’s exciting report from Tango. In an early-stage clinical trial, patients with advanced pancreatic cancer benefited more from a combination of two targeted drugs — a PRMT5 inhibitor from Tango and Revolution’s pan-RAS inhibitor — than they might from each drug on its own.

Continue to STAT+ to read the full story…

Posted on

Evaluating Postpartum Hemorrhage Transfusion Risk With a Machine Learning Model for Informed Consent: Retrospective Cohort Study

<strong>Background:</strong> Postpartum hemorrhage requiring a blood transfusion is a concern for patients and clinicians; its risk and the mode of delivery are important points of discussion before labor. Many high-risk factors associated with postpartum hemorrhage are known prior to delivery and are often unpreventable. Delivery plans are influenced by the patient’s medical history, their preferences, and clinical decision-making. Informed consent regarding known risk factors for postpartum hemorrhage will help guide delivery care plans and mitigate risk. Machine learning models have been used to predict postpartum hemorrhage; however, translation into clinical support tools is challenging. Shared decision-making discussions can be facilitated with machine learning model–based clinical support tools predicting postpartum hemorrhage requiring a transfusion. <strong>Objective:</strong> We sought to develop a machine learning model for prediction of postpartum hemorrhage requiring a transfusion. Specifically, we sought to evaluate the model’s accuracy in predicting a patient’s postpartum transfusion risk based on delivery mode, whether labor was induced, and the delivery indications for the purpose of antenatal clinical decision support. Model performance was evaluated on existing structured data and physician-reviewed datasets. <strong>Methods:</strong> A 10-year retrospective cohort of 62,521 births in a community health system was sampled. A convenience sample of 1734 patients was analyzed to predict blood transfusion rate based on delivery mode and delivery indications. XGBoost, random forest, and generalized linear models were trained and compared for performance. Datasets were evaluated using the best-performing XGBoost machine learning model. A prototype clinical support app for physician-patient transfusion risk assessment was developed using the best-performing clinically relevant XGBoost model. <strong>Results:</strong> A generalized linear model, random forest model, and XGBoost model were evaluated. The XGBoost model was trained with an existing dataset extracted from electronic medical records (n=1734). The area under the curve (AUC) was 0.71, precision-recall receiver operating characteristic curve (PR-ROC) was 0.82, and <i>F</i><sub>1</sub> score was 0.80. Performance on a physician-reviewed dataset (n=1734) was as follows: AUC=0.705, PR-ROC=0.78, and <i>F</i><sub>1</sub> score=0.809. Feature importance ranking and prediction were not clinically accurate for the pre-review dataset. <strong>Conclusions:</strong> Machine learning models are useful to determine an individual’s postpartum transfusion risk based on clinically variable and potentially modifiable factors, such as delivery mode, whether labor is induced, and delivery indications. In this study, the XGBoost model had a slightly higher AUC on structured data extracted from electronic medical records than the same dataset after physician review (AUC=0.71 and PR-ROC=0.82 vs AUC=0.705 and PR-ROC=0.78), but a slightly lower <i>F</i><sub>1</sub> score (<i>F</i><sub>1</sub>=0.80 vs <i>F</i><sub>1</sub>=0.809). XGBoost machine learning models trained on clinician-reviewed data can be used to predict postpartum transfusion. Clinically relevant, physician-labeled datasets are important for supervised machine learning model training for use in clinical decision support tools. Further study and external validation are needed prior to clinical use.
Posted on

Automated Machine Learning Frameworks for Radiomics: Comparative Evaluation Study

<strong>Background:</strong> Automated machine learning (AutoML) frameworks can lower technical barriers for predictive and prognostic model development in radiomics by enabling researchers without programming expertise to build models. However, their effectiveness in addressing radiomics-specific challenges remains unclear. <strong>Objective:</strong> This study aimed to evaluate the performance, efficiency, and accessibility of general-purpose and radiomics-specific AutoML frameworks on diverse radiomics classification tasks, thereby guiding researchers and highlighting development needs for radiomics. <strong>Methods:</strong> A total of 10 public and private radiomics datasets with varied imaging modalities (computed tomography and magnetic resonance imaging), sizes, anatomies, and end points were used. Six general-purpose and 5 radiomics-specific frameworks were tested with predefined parameters using standardized cross-validation. Evaluation metrics included area under the receiver operating characteristic curve, runtime, and qualitative aspects related to software status, accessibility, and interpretability. <strong>Results:</strong> Simplatab, a radiomics-specific tool with a no-code interface, achieved the best overall balance between performance and computational efficiency, recording the highest average test area under the receiver operating characteristic curve (mean 78.46%, SD 12.22%) with a moderate runtime (1.1 h). However, its performance was not statistically superior to the most intensive general-purpose solutions. Most radiomics-specific frameworks were excluded from the performance analysis due to obsolescence, extensive programming requirements, or computational inefficiency. Conversely, general-purpose frameworks demonstrated higher accessibility and ease of implementation. <strong>Conclusions:</strong> While no single framework demonstrated absolute predictive superiority, Simplatab provides an effective balance of performance, efficiency, and accessibility for radiomics classification problems. However, continued efforts are needed to further mature AutoML solutions in the radiomics domain.
Posted on

Educational Intervention on Environmentally Responsible Inhaler Prescribing Among French General Practitioners: Pilot Pre-Post Study

<strong>Background:</strong> Climate change is expected to cause more than 250,000 deaths annually by 2050 and could increase the prevalence of asthma and chronic obstructive pulmonary disease (COPD) by up to 30%. Pressurized metered-dose inhalers (pMDIs), primarily delivering short-acting beta-2 agonists, generate 15 to 30 times more greenhouse gas emissions than dry powder or soft mist inhalers. In France, short-acting beta-2 agonist pMDIs account for 95% of reliever therapy prescriptions, despite their limited effectiveness in controlling disease symptoms. <strong>Objective:</strong> This study aimed to evaluate the preliminary educational impact of a single educational session on French general practitioners’ awareness and intended prescribing of lower-carbon inhaler alternatives. <strong>Methods:</strong> We conducted a multicenter, single-group pre-post pilot study among 34 general practitioners from 10 multiprofessional health centers in Eastern Occitanie, France, between March and October 2023. Participants were recruited through convenience sampling. The intervention consisted of a one-time 25-minute face-to-face educational session on environmentally responsible inhaler prescribing, aligned with Global Initiative for Asthma (GINA) and Global Initiative for Chronic Obstructive Lung Disease guidelines. Data were collected using self-administered online questionnaires before the intervention and approximately 3 months later. The questionnaires included 2 clinical vignettes, one on asthma and one on COPD, with 3 prescribing questions each. Responses were categorized according to whether they included a pMDI. Changes in responses between baseline and follow-up were analyzed using the Fisher exact test or chi-square test, as appropriate. <strong>Results:</strong> A total of 34 participants completed the baseline questionnaire. Responses including a pMDI decreased from 70.6% (48/68) to 4% (3/68) for reliever therapy (<i>P</i>&lt;.001) and from 21.3% (29/136) to 4.4% (6/136) for maintenance therapy (<i>P</i>=.003). In asthma scenarios, adherence to GINA recommendations improved, with increased responses including inhaled corticosteroid-formoterol for reliever therapy (6%, 2/34 to 38%, 13/34; <i>P</i>=.001) and maintenance therapy (35%, 24/68 to 56%, 38/68; <i>P</i>=.02). No significant improvements were observed for COPD-related prescribing scenarios. The proportion of participants reporting environmental impact as a factor influencing inhaler choice increased from 3% (1/34) to 51% (18/34). Satisfaction was high, with 93% of participants reporting being very satisfied with the intervention. <strong>Conclusions:</strong> This pilot study suggests that a brief educational intervention may improve general practitioners’ knowledge and intended prescribing of lower-carbon inhaler alternatives, particularly in asthma scenarios. However, the outcomes were based on theoretical clinical vignettes rather than real-world prescribing data, and the study was not designed to assess the safety or clinical effectiveness of changing inhaler prescriptions. Future studies should evaluate sustained changes in real-world prescribing while ensuring individualized, clinically appropriate, and safe inhaler choices.
Posted on

Google DeepMind is worried about what happens when millions of agents start to interact

Google DeepMind is funding research into the potential dangers of situations where millions of different AI agents interact with each other online.

According to Rohin Shah, who directs the company’s AGI safety and alignment research, the mass-market arrival of agents that can carry out tasks without human oversight and follow instructions given to them by other agents creates a whole new class of risk.

In an effort to address this, Google DeepMind—which made agent-based tools a centerpiece of Google I/O last month—has teamed up with several other organizations to announce a $10 million funding pot for researchers to study the behavior of multi-agent systems and come up with ways to prevent unsafe scenarios. Joining Google DeepMind are Schmidt Sciences, a philanthropic foundation set up by Eric and Wendy Schmidt; ARIA, the UK government’s moonshot agency; the Cooperative AI foundation, a UK-based nonprofit research outfit; and Google’s charitable arm, Google.org.

I asked Shah and James Fox, who leads the Science of Trustworthy AI program at Schmidt Sciences, what they hope to achieve with that $10 million. It’s no small sum, but it’s dwarfed by the budgets commanded by Google DeepMind’s own research teams.

The aim is to kick-start research outside tech companies, says Shah: “The strength of academia is that it can look really quite far into the future and do the kind of work that isn’t top of mind at industry labs.”

“The main issue is that there just isn’t really a field of research for multi-agent safety yet,” he adds. “And we would like there to be.”

The concern is that as more and more AI agents get deployed and begin working together, we could hit a tipping point where imagined scenarios become real. “We see this with humanity, too,” says Shah. “Our institutions can accomplish things that no individual human can.”

Shah thinks that we have a few more months to go before agents are deployed throughout the economy in numbers that make potential risks a real concern. He wants to get ahead of that moment.

Risky business

What risks are we talking about, exactly? The possibilities that Shah and Fox have in mind mostly boil down to supercharged versions of bad things that happen on the internet already: scams, prompt injections (where an AI agent is fed malicious instructions, turning it into a self-guiding piece of malware), other forms of cyberattack. We look at what humans do now and ask what the agent version of that would be, says Shah.  

“We’ve got this digital commons that is integral to how society works, and you really want to ensure that this doesn’t descend into just absolute anarchy,” says Fox.

(I asked Shah if they were considering any worst-case scenarios more on the doomer end of the spectrum, such as widespread economic collapse. “Certainly not if we’re talking by the end of the year,” he said. That’s only six months away! He laughed. “Okay, a while after that.”)

Shah and Fox both think that the only way to understand what might happen when large numbers of multi-agent systems interact with each other is to run realistic simulations. They want researchers to drop AI agents into sandboxes and study what they do.

You can’t predict what’s going to happen by studying single agents, or even small groups of agents, in isolation. You can’t assume that AI agents underpinned by LLMs will always act rationally, says Fox. And the complexity comes from having huge numbers of interactions at once.

Some researchers, including a team at Google DeepMind, have argued that artificial general intelligence (if possible at all) could come not from a single super-smart model but from a kind of agent hive mind, where the capabilities of the whole add up to more than the sum of its parts.  

Lack of trust

Google DeepMind is not the only top AI firm warning about the risks of the technology it is building. A couple of weeks ago, Anthropic published guidelines for deploying AI agents based on an approach to cybersecurity known as zero trust, which starts with the assumption that a computer system is vulnerable, an agent is an attacker, and a breach will happen.

Refael Angel, cofounder and CTO of Akeyless, a cybersecurity firm based in Tel Aviv, agrees that understanding the new risks introduced by agent-based systems is crucial.  

Every approach to security in the past has assumed that the machine in question was software written by a human, doing fixed things on fixed paths, says Angel: “An agent breaks all of those assumptions. It reasons, it improvises, and it can be hijacked by a single sentence buried in a document it was asked to read.”

Angel welcomes this new funding. “No single lab should author the safety standards everyone else has to trust,” he says. But he cautions that safety researchers can overlook boring problems that are already here in favor of more exotic hypothetical ones.

And yet, Fox notes, risks that were hypothetical a few years ago are now very real: “The future’s come more quickly than perhaps expected.”