PURPOSE: To determine the accuracy of large language models (LLMs) in generating appropriate treatment options for patients with early breast cancer on the basis of their medical records. MATERIALS AND METHODS: Retrospective study using anonymized medical records of patients with BC presented during multidisciplinary team meetings (MDTs) between January and April 2024. Three generalist artificial intelligence models (Claude3-Opus, GPT4-Turbo, and LLaMa3-70B) were used to generate treatment suggestions, which were compared with experts' decisions. The primary outcome was the rate of appropriate suggestions from the LLMs, compared with the reference experts' decisions. The secondary outcome was the LLMs' performances (F1 score and specificity) in generating appropriate suggestions for each treatment category. RESULTS: The rates of appropriate suggestions were 86.6% (97/112), 85.7% (96/112), and 75.0% (84/112) for Claude3-Opus, GPT4-Turbo, and LLaMa3-70B, respectively. No significant difference was found between Claude3-Opus and GPT4-Turbo ( P = .85), but both tended to perform better than LLaMa3-70B ( P = .027 and P = .043, respectively). LLMs showed high accuracy for adjuvant endocrine therapy and targeted therapy indications. However, they tended to overestimate the need for adjuvant radiotherapy and had variable performances in suggesting adjuvant chemotherapy and genomic tests. CONCLUSION: LLMs, particularly Claude3-Opus and GPT4-Turbo, demonstrated promising accuracy in suggesting appropriate adjuvant treatments for patients with early BC on the basis of their medical records. Although LLMs showed limitations in validating surgery and indicating genomic tests, their performance in other treatment modalities highlights their potential to automate and augment decision making during MDTs. Further studies with fine-tuned LLMs and a prospective design are needed to demonstrate their utility in clinical practice.
Prostate cancer recurrence in patients previously treated with radical prostatectomy and radiation therapy is challenging. Re-irradiation could be an option, but data regarding efficacy and safety are lacking. We retrospectively evaluated salvage re-irradiation for local recurrence after prostatectomy and external beam radiation therapy. We collected data from 48 patients who underwent salvage reirradiation with stereotactic radiation therapy for local prostate cancer recurrence in the prostatic bed at four French centers. Fifteen patients (31%) were on androgen deprivation therapy during stereotactic radiotherapy. Biochemical response and relapse-free survival were analyzed, and post-treatment toxicities were assessed according to the Common Terminology of Adverse Events criteria. Five patients had grade 3 late bladder toxicity (cystitis), three had grade 3 late incontinence, and one had grade 3 late chronic pain. At three months, 83% of patients had a positive biochemical response. The median follow-up was 22 months. At the end of the follow-up, 21 patients (43%) had a biochemical relapse. The median time to biologic relapse was 27 months. The biochemical relapse rates at 1 and 2 years were 80% and 52%, respectively. In conclusion, salvage re-irradiation for recurrent prostate cancer in the prostate bed may generate significant toxicity rates, and a prospective study with appropriate patient selection is needed to evaluate its effectiveness.
22 February 2023 in Clinical & Experimental Allergy
Non-invasive ventilation (NIV) improves survival of patients with chronic respiratory failure (CRF). Most often, pressure settings are made to normalize arterial blood gases. However, this objective is not always achieved due to intolerance to increased pressure or poor compliance. Few studies have assessed the effect of persistent hypercapnia on ventilated patients’ survival. Data from the Pays de la Loire Respiratory Health Research Institute cohort were analyzed to answer this question. NIV-treated adults enrolled between 2009 and 2019 were divided into 5 subgroups: obesity-hypoventilation syndrome (OHS), COPD, obese COPD, neuromuscular disease (NMD) and chest wall disease (CWD). PaCO2 correction was defined as the achievement of a PaCO2 < 6 kPa or a 20% decrease in baseline PaCO₂ in COPD patients. The endpoint was all-cause mortality. Follow-up was censored in case of NIV discontinuation. Data from 431 patients were analyzed. Median survival was 103 months and 148 patients died. Overall, PaCO2 correction was achieved in 74% of patients. Bivariate analysis did not show any survival difference between patients who achievedPaCO₂ correction and those who remained hypercapnic: overall population: p = 0.74; COPD: p = 0.97; obese COPD: p = 0.28; OHS: p = 0.93; NMD: p = 0.84; CWD: p = 0.28. Moderate residual hypercapnia under NIV does not negatively impact survival in CRF patients. In individuals with poor tolerance of pressure increases, residual hypercapnia can therefore be tolerated under long-term NIV. Larger studies, especially with a higher number of patients with residual PaCO2 > 7 kPa, are needed to confirm these results.
PURPOSE: Medical free texts such as pathology reports contain valuable clinical data but are challenging to structure at scale. Traditional natural language processing approaches require extensive annotated data and training. We investigate the use of large language model (LLM) like Mistral to automatically extract three breast cancer (BC) biomarkers from pathology reports. MATERIALS AND METHODS: We developed and evaluated a pipeline combining Mistral Large LLM and a postprocessing phase. The pipeline's performance was assessed both at document and patient levels. For evaluation, two data sets were used: a data set of 1,152 pathology reports associated with 150 patients with BC focused solely on biomarker values and a gold standard database containing 101 patients with metastatic BC, enriched with detailed patient and tumor characteristics and double-blind validated by clinical research assistants. We also explored the pipeline's performance according to the use of a confidence prompt (CP), a chain of thought (CoT), and few-shot examples. RESULTS: Our extraction pipeline achieved F1 scores of more than 95% and both recall and precision of more than 94% for each biomarker of interest (ie, estrogen receptor, progesterone receptor and human epidermal growth factor receptor 2 status and score) at the document level. At the patient level, the F1 score decreased between 87% and 90% with a greater drop in recall (ranging between 83% and 87%) compared with precision, which remained >90%. The results were similar whether the pipeline included a CP, CoT, or few-shot examples. CONCLUSION: Our study provides strong evidence of the potential of LLMs like Mistral Large for extracting structured BC biomarker data from pathology reports and the potential of such methods for broader digital transformation of health care documents.