![\"Contrastive](\"./training-a-model-that-understands-your-notes-7x-better-than-openai/_assets/finetuning-architecture.avif\")
import Dateline from '../../components/blog/Dateline.astro';
\nSan Francisco, California It's no secret that ML and AI is going to transform healthcare, but building the models to do so is incredibly difficult. Unlike other \"easier\" domains, such as self driving cars or coding, where there are clear constraints on the action space and the input signal is (almost) fully observable, the medical domain is unbounded and only partially observed. Here, the signal about a patient is severely undersampled in the sense that we only extract certain observations about what is actually happening internally to them, such as blood tests, vitals, etc., but ignore a whole host of information, such as genomics, detailed logs of the patient history (did they live with asbestos in 1967?) etc. An analogy to the \"easier\" domains would be like training a self driving car that can only take an image of a constrained view, or drive without knowing the speed or direction it's going for most of the time; or training a coding agent without looking at full code files. Which, as you can probably understand, would be incredibly difficult.
\nSo if this is so hard, why even bother? Because we believe that this is where the value of AI really lies, and not in replacing low skilled workers, which is what is typically happening. Medical predictions is a much harder problem, because as innovators it requires us to reimagine what is physically possible without using the guiding light of what humans can already do to show us the way; and secondly for the reason mentioned above, there often simply isn't the data in a format that is amenable to modelling right now, but we're getting there, and there are now methods to extract data that we can use to great effect.
\nSomething that has enabled a step change towards this is the advent of LLMs, which now enables the large scale feature extraction of key events and values in a patient's clinical notes. Is this perfect? No, as information is not always in the notes, and we need to know a priori what data to use. But it enables us to now leverage predictive features that up to this point were not available at this scale. A typical outcome from a patient event extraction would look something like the data below, and can be extracted for less than $0.01.
\nEVENT_DATE,EVENT_TOKEN\n1972-07-03 00:00:00,DEMO_SEX_MALE\n2018-03-12 09:00:00,ADM_Outpatient\n2018-03-12 00:00:00,DIAG_I10\n2019-11-01 00:00:00,DIAG_R91\n2019-11-01 00:00:00,LAB_PDL1_POSITIVE_60PCT\n2019-11-01 00:00:00,LAB_EGFR_L858R\n2023-01-15 08:00:00,PROC_BRONCHOSCOPY\n2023-01-15 00:00:00,PATHOLOGY_NSCLC_ADENOCARCINOMA\n2023-01-15 14:23:00,ADM_Routine/Elective\n2023-02-10 00:00:00,MED_TYROSINE_KINASE_INHIBITOR\n2023-02-10 07:10:00,ADM_Start_Treatment\n2023-06-05 00:00:00,DIAG_R63\n2023-06-05 00:00:00,MED_ANTIEMETIC\n2023-06-05 10:44:00,ADM_Followup\n2024-03-22 00:00:00,PROC_LOBECTOMY_RIGHT\n2024-03-22 06:50:00,ADM_Surgery\n2024-09-14 00:00:00,LAB_TMB_HIGH\n2024-09-14 00:00:00,MED_IMMUNOTHERAPY\n2024-09-14 09:00:00,ADM_Infusion_Center\n2024-09-18 09:00:00,CENSORED\nThis is simulated data, not PHI.
\nSo how do we model this? Sequence modelling problems have been worked on for a very long time (with moderate effect) until a very popular method called the transformer emerged in 2018. The transformer is nearly 8 years old now, but despite this, its impact in healthcare is only just starting to be felt. At its heart, it's a fantastic architecture for modelling autoregressive sequences (of which language is particularly well suited), but perhaps what is of greater impact is modelling the sequences of key medical events that occur over a patient's lifespan.
\nIt's the combination of large scale data extraction, coupled with the computational power of these models, that enables a new era of predictive modelling in patient outcomes. Why is the transformer such a powerful model? Because it is excellent at dynamically modelling pairwise events and then systematically composing them to create complex interdependence between several events. We're starting to see models such as Delphi-2m and SleepFM show impressive performance at predicting outcomes such as mortality or disease incidence. What is more powerful though, is these models' natural ability to predict a range of potential outcomes and not just single events such as mortality.
\nThis is a big swing we're taking at Radical, where we're building a single unified model that models the entire history of oncology patients by encapsulating the relationships between key events during a patient's lifespan, and then simulating what outcomes for the patient look like. Doing so is incredibly valuable, as it allows patients to make more informed decisions about their care, provides alerts for when patients should seek extra testing, and any other medical outcomes that are being directly modelled in the data. This is an inflection point for what's becoming possible: the architectures are now becoming powerful enough to model these long range dependencies and compose them into robust predictions, and most essentially hospitals are opening up their data silos to enable these predictions to happen. At Radical we're partnering with top US centres to provide the data of 11m patients and billions of temporal events to power our unified Aesclea model.
\nThe model is still cooking, and we'll provide some results very soon, so watch this space. If you're interested in what we're building, want to brainstorm or simply have a chat, please reach out, we'd be really keen to talk to you.
\n", "date_published": "2026-01-03T00:00:00.000Z", "image": "https://radicalhealth.ai/og/blog-aesclea-our-oncology-foundation-model.png", "tags": [ "research" ] }, { "id": "https://radicalhealth.ai/blog/radical-health-perfect-100-on-medical-licensing-exam/", "url": "https://radicalhealth.ai/blog/radical-health-perfect-100-on-medical-licensing-exam/", "title": "Radical Health's AI Achieves 100% on the US Medical Licensing Exam", "summary": "Radical Health's AI scored a perfect 100% on the USMLE, validation of the knowledge engine powering personalized cancer reports for patients and doctors.", "content_html": "import Dateline from '../../components/blog/Dateline.astro';
\nWe're validating a new standard of medical knowledge to ensure no cancer patient has to wonder, \"Is there a better option I don't know about?\"
\nSan Francisco, California Today, we are thrilled to announce a landmark achievement: Radical Health's AI has earned a perfect 100% score on the United States Medical Licensing Examination (USMLE), a comprehensive exam required for all physicians to practice medicine in the US.
\nOur AI recently demonstrated mastery of medical knowledge, far exceeding the 60% score physicians typically need to pass the rigorous exam (access the full report here). This achievement isn't about replacing the crucial role of your oncologist; it's about augmenting their expertise. We see this as a pivotal step toward a future where every patient, in partnership with their doctor, can benefit from the full scope of medical knowledge when making life's most important decisions.
\nIf you or a loved one has been diagnosed with cancer, you know the feeling of uncertainty that follows. The late-night research, the complex medical terms, and the nagging fear that you might be missing a critical piece of the puzzle. Knowing the right questions to ask your oncologist helps; having a system that has memorized every answer to those questions, and the latest research behind them, helps more.
\nThat fear is not unfounded. A landmark study from the Mayo Clinic revealed that up to 88% of patients who sought a second opinion with them received a new or refined diagnosis changing their care plan or lives.
\nExpertise matters profoundly. Yet, access to these leading centers can take months and cost tens of thousands of dollars, leaving the vast majority of patients to navigate their journey without this crucial perspective.
\nEven the best doctors, however brilliant, cannot be expected to have instantaneous recall of the latest research, every new clinical trial, rare drug interaction, and genomic markers and their relevance to your personal medical case. Human physicians provide something irreplaceable: wisdom, empathy, and clinical judgment. Our goal is to arm them, and you, with a tool that masters the other half of the equation: comprehensive, up-to-the-minute data placed in your context.
\nOur 100% score on the USMLE is more than a statistic. It is validation, proving the rock-solid foundation of part of the knowledge engine that powers our platform.
\nBuilt in collaboration with experts from world-leading institutions like UCSF and the Mayo Clinic, the Radical Healthcare platform acts as your personal cancer specialist AI, in your corner, available 24/7.
\nHere's how it empowers you and your care team:
\nThink of it as having a dedicated team of researchers working 24/7 on your specific case, delivering a concise briefing to you and your doctor before your next appointment.
\nOur AI's 100% score on the USMLE is testament to its robust knowledge engine. The specific model that achieved this result is our knowledge recall module, forming part of our larger architecture. We achieved this performance largely through a novel medical-focused multi-agent architecture with tuned real-time prioritized search and grounding.
\nThe model was evaluated on an official sample USMLE exam totaling 325 questions prepared by Kung et al., 2023 and publicly available at usmle.org. We made two specific edits to the data used for evaluation: we identify and correct an erroneously recorded answer for Question #78 from the Step 3 examination. Similarly, Question #125 from the Step 3 exam has an incorrect answer, which we correct.
\nWe envision a future where every cancer patient, regardless of their location or background, has access to world-class oncological expertise.
\nThe 100% USMLE score is an early milestone, not the destination. It is the validation of our commitment to building a trusted, reliable, and accessible source of medical knowledge. The real goal is to ensure that every patient can confidently say, \"I know all my options. This one is the best.\"
\nIf you or a loved one is facing a cancer diagnosis, you deserve to know all your options. Because when it comes to cancer, you shouldn't have to settle for uncertainty. Get your AI-powered personalized cancer report in hours at platform.radicalhealth.ai.
\nFor healthcare providers, researchers, or investors interested in partnering with us, contact us at contact@radicalhealth.ai.
\nFootnote: We recognize that benchmarks like the USMLE, while important, don't fully capture what cancer patients actually need: comprehensive treatment evaluation, personalized therapy selection, and synthesis of cutting-edge research into actionable guidance. Our platform is powered by a system of AI agents that do much more than answer questions, and we're working to soon publish new benchmarks that better reflect real-world oncology care.
\n", "date_published": "2025-09-20T00:00:00.000Z", "image": "https://radicalhealth.ai/og/blog-radical-health-perfect-100-on-medical-licensing-exam.png", "tags": [ "company" ] }, { "id": "https://radicalhealth.ai/blog/cancer-clinical-trials-explained-part-2/", "url": "https://radicalhealth.ai/blog/cancer-clinical-trials-explained-part-2/", "title": "Cancer Clinical Trials Explained: Should You Participate? (Part 2)", "summary": "Cancer clinical trials, the decision: eligibility, consent, what to ask, and what to do if you don't qualify for any trial.", "content_html": "This is Part 2 of our clinical trials guide. If you haven't read it yet, Part 1 covers what trials are, the common myths, the four phases, and the trade-offs. This part focuses on the decision: eligibility, the questions to ask, and what to do if you don't qualify.
\nEvery trial has strict inclusion criteria (what you must have) and exclusion criteria (what disqualifies you).
\nFrustration alert: You might match 90% of criteria but be excluded for one factor.
\nWhy so strict?
\nImportant: Criteria sometimes relax as trials progress and more safety data is available.
\nBefore enrolling, you'll go through informed consent.
\nWritten consent form (often 20 to 30 pages) explaining:
\nVerbal explanation from:
\nTime to consider (never rush):
\nImportant: Signing consent is NOT a final commitment. You can withdraw at any time.
\nConsider \"Yes\" if:
\nConsider \"No\" or \"Not yet\" if:
\nMany patients benefit from discussing the trial with:
\nRemember: This is YOUR decision. No one should pressure you either way.
\nRegular monitoring:
\nStrict adherence required:
\nCommunication:
\nMany patients are disappointed to find they don't meet eligibility criteria. Options:
\nWhat they are: Trials that accept multiple cancer types sharing a specific genetic mutation.
\nExample:
\nAdvantage: Expands options for rare mutations regardless of cancer type.
\nWhat they are: Multiple sub-studies under one master protocol.
\nHow it works:
\nExamples: NCI-MATCH, Lung-MAP.
\nAdvantage: Personalized matching to treatment based on your tumor's specific characteristics.
\nWhat they are: Brief treatment before planned surgery.
\nPurpose:
\nAdvantage: Contribute to research without changing your standard treatment plan.
\nWhat they are: Treatment BEFORE surgery (instead of or in addition to standard neoadjuvant therapy).
\nPurpose:
\nExample: Immunotherapy before surgery for certain cancers.
\nClinical trials are how cancer care advances. Every effective treatment available today started in a clinical trial.
\nShould you participate? Consider it if:
\nIt's also perfectly okay to choose standard treatment if:
\nCancer treatment is increasingly personalized. Clinical trials are one way to access the most personalized, cutting-edge approaches, but they're not the only path to excellent care.
\nimport Disclaimer from '../../components/blog/Disclaimer.astro';
\nimport Dateline from '../../components/blog/Dateline.astro';
\nSan Francisco, California Our new medical EmbeddingGemma-300m fine-tuned on medical data can predict diseases from patient notes with an AUROC of 0.934 and provide increased survival analysis.
\nThe current problem with embedding models is that they are trained to match text content, but text covers a whole range of information, from language to style, to semantics, to the underlying information or instructions. When it comes to needing embeddings which contain the specific information and are agnostic to stylistic information, off-the-shelf embeddings typically fail.
\nFor the medical context this is critical, as we need to be able to faithfully extract the correct clinical information and prioritise this over other stylistic information.
\nWe built the first generation of an internal embedding model by fine-tuning the EmbeddingGemma-300m using the MIMIC-III dataset, which is a large, freely available critical care database containing de-identified health records from over 40,000 ICU patients between 2001 and 2012. Using the contrastive loss, we fine-tune the embedding model by setting the anchor to be a patient note at time t and the positive to be a patient note at t + 1. This forces the model to match notes based on the medical context and not to rely on the style of writing, which is typically consistent between notes.
\n
Fine-tuning a model this way yields surprisingly powerful results. Given a recall task of recalling the next patient note, our model is able to achieve a top-5 accuracy of 65%, far surpassing the accuracy of the base model and OpenAI embeddings, which achieve 6% and 9% respectively.
\nThis isn't surprising as we're optimising for this in the loss, but where this gets exciting is that these representations make for much better medical performance on downstream tasks. Our model beats the base model and OpenAI when trying to predict diagnosis (ours: 0.934, OpenAI: 0.809, base: 0.674 AUROC), and when performing survival analysis (ours: 0.70, OpenAI: 0.67, base: 0.59 C-Index).
\nWe can clearly see why the model is so powerful at producing these results when we create UMAP plots and color code by common disease type. Compared against the off-the-shelf base model and OpenAI embeddings, our trained model produces visibly tighter clusters per diagnosis:
\n![\"UMAP](\"./training-a-model-that-understands-your-notes-7x-better-than-openai/_assets/umap-base-model.avif\")
![\"UMAP](\"./training-a-model-that-understands-your-notes-7x-better-than-openai/_assets/umap-openai-model.avif\")
![\"UMAP](\"./training-a-model-that-understands-your-notes-7x-better-than-openai/_assets/umap-trained-model.avif\")
The contrast tells the story: the same clinical notes, the same diagnoses, but a clinically meaningful geometry emerges only after fine-tuning. This is the foundation we're building toward Aesclea, our long-range temporal model of oncology patients, and the same medical-knowledge stack that lets our patient-report platform score 100% on the US Medical Licensing Exam. See how it works end-to-end for the patient-facing product.
\nIf you're working on medical AI and want to compare notes (or use these representations on a real downstream task), reach out.
\n", "date_published": "2025-04-13T00:00:00.000Z", "image": "https://radicalhealth.ai/og/blog-training-a-model-that-understands-your-notes-7x-better-than-openai.png", "tags": [ "research" ] }, { "id": "https://radicalhealth.ai/blog/cancer-clinical-trials-explained-part-1/", "url": "https://radicalhealth.ai/blog/cancer-clinical-trials-explained-part-1/", "title": "Cancer Clinical Trials Explained: Should You Participate? (Part 1)", "summary": "Cancer clinical trials, the basics: phases, common myths, types of trials, and how participation fits into modern cancer care.", "content_html": "When you hear \"clinical trial,\" you might think it's a last resort for patients who've run out of options. This is one of the biggest misconceptions in cancer care.
\nThe reality: clinical trials are often where the best, most cutting-edge treatments are available, sometimes years before they become standard care. Many trials are specifically for newly diagnosed patients, and participation in clinical trials is associated with better outcomes.
\nThis comprehensive guide explains everything you need to know about cancer clinical trials to make an informed decision. It comes in two parts; you're reading Part 1, which covers the basics. Part 2 covers eligibility, the questions to ask, and the consent process. For a broader checklist of what to bring up at your next oncology appointment, see the five questions every cancer patient should ask their oncologist.
\nA clinical trial is a carefully controlled research study that tests:
\nThe goal: determine if new approaches are:
\nEvery cancer treatment you have access to today was once in a clinical trial:
\nPatients who participated in these trials often had better outcomes than those receiving standard care at the time.
\nReality: Many trials are ONLY for newly diagnosed or early-stage patients.
\nExamples:
\nFact: Some of the most promising trials specifically exclude patients who've had previous treatment.
\nReality: In cancer trials, you rarely receive a placebo alone.
\nHow it actually works:
\nImportant: If you're in a trial comparing Treatment A vs. Treatment B, you'll receive one of them, both are real treatments.
\nReality: Clinical trials have extensive safety oversight.
\nSafety measures include:
\nFact: Clinical trial participants often receive MORE careful monitoring than patients receiving standard care.
\nReality: By the time a treatment reaches human trials, it has undergone extensive laboratory and animal testing.
\nThe development process:
\nYou participate in Phase 1, 2, or 3, after years of preliminary research.
\nReality: It's complicated.
\nWhat's typically covered by the trial:
\nWhat you/insurance typically pay:
\nImportant: The Affordable Care Act requires most insurance to cover routine costs of clinical trial participation.
\nUnderstanding trial phases helps you assess potential benefits and risks.
\nPurpose: Determine safe dosage and identify side effects
\nParticipants: 20 to 80 people, usually with advanced cancer
\nWhat happens:
\nSuccess rate: ~5% of participants have tumor shrinkage (not the goal; safety is the goal)
\nConsider Phase 1 if:
\nRisks: Unknown side effects, may not help your cancer
\nBenefits:
\nPurpose: Determine if the treatment works against specific cancers
\nParticipants: 100 to 300 people with specific cancer types
\nWhat happens:
\nSuccess rate: ~30% of Phase 2 trials show enough promise to move to Phase 3
\nResponse rates: Vary widely (10 to 70% of participants might have tumor response)
\nConsider Phase 2 if:
\nRisks: Treatment may not work, side effects still being characterized
\nBenefits:
\nPurpose: Compare new treatment to current standard of care
\nParticipants: 300 to 3,000 people (large trials)
\nWhat happens:
\nSuccess rate: ~33% of Phase 3 trials show new treatment is better
\nConsider Phase 3 if:
\nRisks:
\nBenefits:
\nImportant: Many Phase 3 trials allow \"crossover\", so if you're in the standard arm and the trial shows the new treatment is better, you can switch to it.
\nPurpose: Continue studying a treatment after FDA approval
\nParticipants: Thousands
\nWhat happens:
\nConsider Phase 4 if:
\nTest new:
\nMost common type of cancer trial.
\nTest interventions to prevent cancer in:
\nExamples:
\nTest new ways to detect cancer earlier:
\nExample: Liquid biopsy trials to detect cancer from blood samples.
\nTest interventions to:
\nExamples:
\nPotential medical benefits:
\nData supports this: Studies show clinical trial participants often have better outcomes than similar patients receiving standard care, even those in the \"standard treatment\" arm, likely due to more careful protocol adherence and monitoring.
\nPersonal benefits:
\nFinancial considerations:
\nHowever: you typically still pay for standard care costs.
\nMedical risks:
\nPractical downsides:
\nImportant considerations:
\nContinue to Part 2 for eligibility criteria, the questions to ask before enrolling, the informed consent process, and how to decide.
\nimport BlogFAQ from '../../components/blog/BlogFAQ.astro';\nimport Disclaimer from '../../components/blog/Disclaimer.astro';
\n<BlogFAQ\n heading=\"Common myths about clinical trials, debunked\"\n items={[\n {\n question: \"Are clinical trials only for people who've run out of options?\",\n answer: \"No. Many trials specifically target newly diagnosed patients or those receiving first-line treatment. Trials exist at every stage of the cancer journey, from prevention through advanced disease. Asking about trials early can open access to next-generation treatments that aren't yet standard of care.\",\n },\n {\n question: \"Will I get a placebo and receive no treatment?\",\n answer: \"Placebo-only arms are rare in cancer trials. In most trials, the control arm receives the current standard of care, and the experimental arm receives the standard of care plus the new treatment. Truly inactive placebos are typically only used in supportive-care or prevention trials where no proven standard exists.\",\n },\n {\n question: \"Are clinical trials dangerous experiments?\",\n answer: \"Modern trials go through years of preclinical safety testing and ascending-dose Phase 1 studies before any patient receives the full dose. Trials are continuously monitored by data safety monitoring boards that can stop the trial if harm appears. They are typically safer than off-protocol experimental use because the protections are formalized.\",\n },\n {\n question: \"Will I be treated like a guinea pig?\",\n answer: \"No. Patients in trials often receive more frequent monitoring, more specialist time, and tighter care coordination than patients receiving standard treatment. The informed consent process is extensive, you can withdraw at any time, and your care team's primary obligation remains your individual well-being.\",\n },\n {\n question: \"Are clinical trials free treatment?\",\n answer: \"Partially. The trial sponsor typically covers the experimental drug and trial-specific tests. Standard-of-care costs (routine labs, imaging, hospitalization for unrelated issues) are usually billed to insurance as they would be off-trial. Some trials cover travel and lodging; ask the coordinator for a full cost breakdown before enrolling.\",\n },\n ]}\n/>
\nWhen you're diagnosed with cancer, you'll face more medical decisions in a few weeks than most people make in a lifetime. The difference between good and excellent cancer care often comes down to one thing: asking the right questions.
\nThis guide provides the essential questions every cancer patient should ask, and explains why each answer matters for your care.
\nPreparation tips:
\nYour treatment plan depends entirely on your specific cancer characteristics. \"Breast cancer\" isn't enough, you need details:
\nCancer type examples:
\nStage (0 to IV):
\nGrade (1 to 3 or 4):
\nIf your doctor seems vague about your exact diagnosis or stage, this is a sign you may need a second opinion or additional testing.
\nOncologists sometimes present one recommended treatment without discussing alternatives. You deserve to know:
\nImportant: For many cancers, there are multiple equally effective options. Your oncologist's recommendation may be based on their specialty or institutional preference, not necessarily what's uniquely best for you.
\nModern cancer treatment is increasingly personalized based on your tumor's genetic profile. Testing should happen BEFORE starting treatment when possible.
\nBreast cancer:
\nLung cancer:
\nColon cancer:
\nAll advanced cancers:
\nStarting treatment before complete testing is sometimes necessary (fast-growing cancers), but often patients begin chemotherapy before knowing they might benefit from targeted therapy or immunotherapy with fewer side effects. The breadth of relevant biomarkers, and the speed at which the list grows, is part of why we built an AI knowledge engine that scored 100% on the US Medical Licensing Exam before we let it touch a patient's records.
\nDifferent treatments have different goals:
\nCurative intent:
\nLife extension:
\nPalliative / symptom control:
\nActive surveillance:
\nTreatment side effects can range from minor inconveniences to life-altering complications. You need realistic expectations:
\nCognitive effects (\"chemo brain\"):
\nSexual and fertility effects:
\nAppearance changes:
\nFinancial toxicity:
\nEmotional impact:
\nImportant: Ask for written information about side effects. You won't remember everything discussed in the appointment.
\nimport BlogFAQ from '../../components/blog/BlogFAQ.astro';\nimport Disclaimer from '../../components/blog/Disclaimer.astro';
\n<BlogFAQ\n heading=\"Quick reference: the five questions\"\n items={[\n {\n question: \"What is my exact diagnosis, including the type, stage, and grade of cancer?\",\n answer: \"Ask for the specific cancer type and subtype (not just 'breast cancer' but the histology), the TNM stage, the tumor grade, and the result of any molecular or genetic markers. These details drive every downstream treatment decision and let you research evidence-based options that apply to your case.\",\n },\n {\n question: \"What are ALL my treatment options, including the option of doing nothing?\",\n answer: \"Ask the oncologist to list every reasonable option, including standard care, clinical trials, and watchful waiting. For each, ask about goals (cure vs. control), evidence quality, expected response rate, side effects, and the trade-offs. Request the option to do nothing or wait so you understand the natural history of your disease.\",\n },\n {\n question: \"What genetic testing or biomarker testing should be done on my tumor?\",\n answer: \"Many cancers have actionable biomarkers (EGFR, ALK, BRCA, PD-L1, microsatellite instability, tumor mutational burden, and more) that change which treatments are most likely to work. Ask which tests are recommended for your cancer type and whether results would change the treatment plan or open up clinical-trial options.\",\n },\n {\n question: \"What does success look like, and how will we measure it?\",\n answer: \"Ask about realistic outcomes: cure, remission, stable disease, or symptom control. Ask which imaging or lab values will track progress, how often they'll be repeated, and what would prompt a treatment change. Aligning on the definition of success up front prevents misaligned expectations later.\",\n },\n {\n question: \"What are the side effects, and how will they impact my daily life?\",\n answer: \"Ask for the most common side effects of each treatment, the serious-but-rare ones, how long they typically last, what can be done to manage them, and how they will affect your ability to work, drive, eat, sleep, and care for family. Request written side-effect information you can review at home.\",\n },\n ]}\n/>
\n