Category: Statistics

JAMA Netw Open. 2024 Apr 1;7(4):e244630. doi: 10.1001/jamanetworkopen.2024.4630.

ABSTRACT

IMPORTANCE: Artificial intelligence (AI) large language models (LLMs) demonstrate potential in simulating human-like dialogue. Their efficacy in accurate patient-clinician communication within radiation oncology has yet to be explored.

OBJECTIVE: To determine an LLM’s quality of responses to radiation oncology patient care questions using both domain-specific expertise and domain-agnostic metrics.

DESIGN, SETTING, AND PARTICIPANTS: This cross-sectional study retrieved questions and answers from websites (accessed February 1 to March 20, 2023) affiliated with the National Cancer Institute and the Radiological Society of North America. These questions were used as queries for an AI LLM, ChatGPT version 3.5 (accessed February 20 to April 20, 2023), to prompt LLM-generated responses. Three radiation oncologists and 3 radiation physicists ranked the LLM-generated responses for relative factual correctness, relative completeness, and relative conciseness compared with online expert answers. Statistical analysis was performed from July to October 2023.

MAIN OUTCOMES AND MEASURES: The LLM’s responses were ranked by experts using domain-specific metrics such as relative correctness, conciseness, completeness, and potential harm compared with online expert answers on a 5-point Likert scale. Domain-agnostic metrics encompassing cosine similarity scores, readability scores, word count, lexicon, and syllable counts were computed as independent quality checks for LLM-generated responses.

RESULTS: Of the 115 radiation oncology questions retrieved from 4 professional society websites, the LLM performed the same or better in 108 responses (94%) for relative correctness, 89 responses (77%) for completeness, and 105 responses (91%) for conciseness compared with expert answers. Only 2 LLM responses were ranked as having potential harm. The mean (SD) readability consensus score for expert answers was 10.63 (3.17) vs 13.64 (2.22) for LLM answers (P < .001), indicating 10th grade and college reading levels, respectively. The mean (SD) number of syllables was 327.35 (277.15) for expert vs 376.21 (107.89) for LLM answers (P = .07), the mean (SD) word count was 226.33 (191.92) for expert vs 246.26 (69.36) for LLM answers (P = .27), and the mean (SD) lexicon score was 200.15 (171.28) for expert vs 219.10 (61.59) for LLM answers (P = .24).

CONCLUSIONS AND RELEVANCE: In this cross-sectional study, the LLM generated accurate, comprehensive, and concise responses with minimal risk of harm, using language similar to human experts but at a higher reading level. These findings suggest the LLM’s potential, with some retraining, as a valuable resource for patient queries in radiation oncology and other medical fields.

PMID:38564215 | DOI:10.1001/jamanetworkopen.2024.4630

Transl Vis Sci Technol. 2024 Apr 2;13(4):4. doi: 10.1167/tvst.13.4.4.

ABSTRACT

PURPOSE: Establishing a development environment for machine learning is difficult for medical researchers because learning to code is a major barrier. This study aimed to improve the accuracy of a postoperative vault value prediction model for implantable collamer lens (ICL) sizing using machine learning without coding experience.

METHODS: We used Orange data mining, a recently developed open-source, code-free machine learning tool. This study included eye-pair data from 294 patients from the B&VIIT Eye Center and 26 patients from Kim’s Eye Hospital. The model was developed using OCULUS Pentacam data from the B&VIIT Eye Center and was internally evaluated through 10-fold cross-validation. External validation was performed using data from Kim’s Eye Hospital.

RESULTS: The machine learning model was successfully trained using the data collected without coding. The random forest showed mean absolute errors of 124.8 µm and 152.4 µm for the internal 10-fold cross-validation and the external validation, respectively. For high vault prediction (>750 µm), the random forest showed areas under the curve of 0.725 and 0.760 for the internal and external validation datasets, respectively. The developed model performed better than the classic statistical regression models and the Google no-code platform.

CONCLUSIONS: Applying a no-code machine learning tool to our ICL implantation datasets showed a more accurate prediction of the postoperative vault than the classic regression and Google no-code models.

TRANSLATIONAL RELEVANCE: Because of significant bias in measurements and surgery between clinics, the no-code development of a customized machine learning nomogram will improve the accuracy of ICL implantation.

PMID:38564200 | DOI:10.1167/tvst.13.4.4

Cell Oncol (Dordr). 2024 Apr 2. doi: 10.1007/s13402-024-00940-y. Online ahead of print.

ABSTRACT

INTRODUCTION: Chimeric antigen receptor (CAR)-T cells obtained long-term durability in about 30% to 40% of relapsed/refractory (r/r) B-cell non-Hodgkin lymphoma (B-NHL). Maintenance therapy after CAR-T is necessary, and PD1 inhibitor is one of the important maintenance therapy options.

METHODS: A total of 173 r/r B-NHL patients treated with PD1 inhibitor maintenance following CD19/22 CAR-T therapy alone or combined with autologous hematopoietic stem cell transplantation (ASCT) from March 2019 to July 2022 were assessed for eligibility for two trials. There were 81 patients on PD1 inhibitor maintenance therapy.

RESULTS: In the CD19/22 CAR-T therapy trial, the PD1 inhibitor maintenance group indicated superior objective response rate (ORR) (82.9% vs 60%; P = 0.04) and 2-year progression-free survival (PFS) (59.8% vs 21.3%; P = 0.001) than the non-maintenance group. The estimated 2-year overall survival (OS) was comparable in the two groups (60.1% vs 45.1%; P = 0.112). No difference was observed in the peak expansion levels of CD19 CAR-T and CD22 CAR-T between the two groups. The persistence time of CD19 and CD22 CAR-T in the PD1 inhibitor maintenance group was longer than that in the non-maintenance group. In the CD19/22 CAR-T therapy combined with ASCT trial, no significant differences in ORR (81.4% vs 84.8%; P = 0.67), 2-year PFS (72.3% vs 74.9%; P = 0.73), and 2-year OS (84.1% vs 80.7%; P = 0.79) were observed between non-maintenance and PD1 inhibitor maintenance therapy groups. The peak expansion levels and duration of CD19 and CD22 CAR-T were not statistically different between the two groups. During maintenance treatment with PD1 inhibitor, all adverse events were manageable. In the multivariable analyses, type and R3m were independent predictive factors influencing the OS of r/r B-NHL with PD1 inhibitor maintenance after CAR-T therapy.

CONCLUSION: PD1 inhibitor maintenance following CD19/22 CAR-T therapy obtained superior response and survival in r/r B-NHL, but not in the trial of CD19/22 CAR-T cell therapy combined with ASCT.

PMID:38564164 | DOI:10.1007/s13402-024-00940-y

Hormones (Athens). 2024 Apr 2. doi: 10.1007/s42000-024-00553-y. Online ahead of print.

ABSTRACT

PURPOSE: Evidence from previous experimental and observational research demonstrates that the gut microbiota is related to circulating adipokine concentrations. Nevertheless, the debate as to whether gut microbiome composition causally influences circulating adipokine concentrations remains unresolved. This study aimed to take an essential step in elucidating this issue.

METHODS: We used two-sample Mendelian randomization (MR) to causally analyze genetic variation statistics for gut microbiota and four adipokines (including adiponectin, leptin, soluble leptin receptor [sOB-R], and plasminogen activator inhibitor-1 [PAI-1]) from large-scale genome-wide association studies (GWAS) datasets. A range of sensitivity analyses was also conducted to assess the stability and reliability of the results.

RESULTS: The composite results of the MR and sensitivity analyses revealed 22 significant causal associations. In particular, there is a suggestive causality between the family Clostridiaceae1 (IVW: β = 0.063, P = 0.034), the genus Butyrivibrio (IVW: β = 0.029, P = 0.031), and the family Alcaligenaceae (IVW: β=-0.070, P = 0.014) and adiponectin. Stronger causal effects with leptin were found for the genus Enterorhabdus (IVW: β=-0.073, P = 0.038) and the genus Lachnospiraceae (NK4A136 group) (IVW: β=-0.076, P = 0.01). Eight candidate bacterial groups were found to be associated with sOB-R, with the phylum Firmicutes (IVW: β = 0.235, P = 0.03) and the order Clostridiales (IVW: β = 0.267, P = 0.028) being of more interest. In addition, the genus Roseburia (IVW: β = 0.953, P = 0.022) and the order Lactobacillales (IVW: β=-0.806, P = 0.042) were suggestive of an association with PAI-1.

CONCLUSION: This study reveals a causal relationship between the gut microbiota and circulating adipokines and may help to offer novel insights into the prevention of abnormal concentrations of circulating adipokines and obesity-related diseases.

PMID:38564143 | DOI:10.1007/s42000-024-00553-y

Curr Atheroscler Rep. 2024 Apr 2. doi: 10.1007/s11883-024-01197-4. Online ahead of print.

ABSTRACT

PURPOSE OF REVIEW: Heart failure (HF) is one of the main causes of cardiovascular mortality in the western world. Despite great advances in treatment, recurrence and mortality rates remain high. Soluble guanylate cyclase is an enzyme which, by producing cGMP, is responsible for the effects of vasodilation, reduction of cardiac pre- and after-load and, therefore, the improvement of myocardial performance. Thus, a new therapeutic strategy is represented by the stimulators of soluble guanylate cyclase (sGCs). The aim of this meta-analysis was to analyze the effects deriving from the administration of sGCs, in subjects affected by HF. A systematic literature search of Medline, SCOPUS, and Google Scholar was conducted up to December 2022 to identify RCTs assessing the cardiovascular effects, as NT-pro-BNP values and ejection fraction (EF), and all-cause mortality, of the sGCs. Quantitative data synthesis was performed using a random-effects model, with weighted mean difference (WMD) and 95% confidence interval (CI) as summary statistics.

RECENT FINDINGS: The results obtained documented a statistically significant improvement in NT-proBNP values (SMD: – 0.258; 95% CI: – 0.398, – 0.118; p < 0.001) and EF (WMD: 0.948; 95% CI: 0.485, 1.411; p < 0.001) in subjects treated with sGCs; however, no significant change was found in the all-cause mortality rate (RR 0.96; 95% CI 0.868 to 1.072; I2, p = 0). The sGCs represent a valid therapeutic option in subjects suffering from HF, leading to an improvement in cardiac performance.

PMID:38564140 | DOI:10.1007/s11883-024-01197-4

Environ Sci Pollut Res Int. 2024 Apr 2. doi: 10.1007/s11356-024-33120-4. Online ahead of print.

ABSTRACT

Encouraging the recycling of plastic packaging materials in express delivery is a necessary step toward environmentally friendly industrial development. In this study, we present a framework for analyzing the flow of materials in express plastic packaging, from production and manufacturing to consumption and recycling. In examining the use of recycled materials in post-consumer express plastic packaging and the destination of consumer packaging waste in 2020 and 2021, we found that 44.4% (1613.6 Gg) of the studied express plastic packaging was incinerated. Additionally, approximately 1296.6 Gg of express plastic packaging flowed into rural areas. Our calculations showed that the ΔRSE in 2020 was 15.1%, and on the condition that 25% separated collection with 80% recycling, ΔRSE would be – 0.5%. Results verified that separated collection is an important step in the recycling strategy for packaging materials. Survey data from universities in Beijing indicate that currently, 26% of college students are participating in the separate collection of packaging.

PMID:38564128 | DOI:10.1007/s11356-024-33120-4

J Gastrointest Cancer. 2024 Apr 2. doi: 10.1007/s12029-024-01048-0. Online ahead of print.

ABSTRACT

BACKGROUND: Laparoscopy-assisted gastrectomy (LAG) is a well-established surgical technique in treating patients with early gastric cancer. However, the efficacy and safety of LAG versus open gastrectomy (OG) in patients with advanced gastric cancer (AGC) remains unclear.

METHODS: We systematically searched PubMed, Embase, and Cochrane Library in June 2023 for RCTs comparing LAG versus OG in patients with AGC. We pooled risk ratios (RR) and mean differences (MD) with 95% confidence intervals (CI) for binary and continuous endpoints, respectively. We performed all statistical analyses using R software version 4.3.1 and a random-effects model.

RESULTS: Nine RCTs comprising 3827 patients were included. There were no differences in terms of intraoperative complications (RR 1.14; 95% CI 0.72 to 1.82), number of retrieved lymph nodes (MD -0.54 lymph nodes; 95% CI -1.18 to 0.09), or mortality (RR 0.91; 95% CI 0.30 to 2.83). LAG was associated with a longer operative time (MD 49.28 minutes; 95% CI 30.88 to 67.69), lower intraoperative blood loss (MD -51.24 milliliters; 95% CI -81.41 to -21.06), shorter length of stay (MD -0.83 days; 95% CI -1.60 to -0.06), and higher incidence of pancreatic fistula (RR 2.44; 95% CI 1.08 to 5.50). Postoperatively, LAG was also superior to OG in reducing bleeding rates (RR 0.44; 95% CI 0.22 to 0.86) and time to first flatus (MD -0.27 days; 95% CI -0.47 to -0.07), with comparable results in anastomotic leakage, wound healing issues, major complications, time to ambulation, or time to first liquid intake. In the long-term analyses at 3 and 5 years, there were no significant differences between LAG and OG in terms of overall survival (RR 0.99; 95% CI 0.96 to 1.03) or relapse-free survival (RR 0.99; 95% CI 0.94 to 1.04).

CONCLUSION: This meta-analysis of RCTs suggests that LAG may be an effective and safe alternative to OG for treating AGC; albeit, it may be associated with an increased risk for pancreatic fistula.

PMID:38564116 | DOI:10.1007/s12029-024-01048-0

Int J Clin Oncol. 2024 Apr 2. doi: 10.1007/s10147-024-02489-0. Online ahead of print.

ABSTRACT

BACKGROUND: The present study aimed to examine the effect of magnesium (Mg) supplementation on cisplatin-induced nephrotoxicity (CIN) in pediatric cancer patients.

METHODS: The present phase-2, open-label, multicenter, randomized controlled trial enrolled patients aged less than 20 years who were scheduled to receive cisplatin-containing chemotherapy and randomly allocated them at a ratio of 1:1 to a Mg supplementation arm with even-numbered chemotherapy courses (arm AB) or another arm with odd-numbered courses (arm BA). Analysis objects were reconstructed into two groups depending on whether the chemotherapy course had Mg supplementation (group B) or not (group A). The primary outcome was the proportion of chemotherapy courses resulting in elevated serum creatinine per chemotherapy course. The secondary outcomes included efficacies evaluated using other biomarkers and the safety of the Mg supplementation.

RESULTS: Twenty-eight patients were randomly allocated to either group (16 to arm AB and 12 to arm BA). The baseline characteristics of the groups were similar. There was no significant difference in the proportion of courses with increased serum creatinine between the groups (group A: 10% vs. group B: 6%; P = 0.465) nor was any significant difference observed in other biomarkers during any chemotherapy course. The Mg value during chemotherapy was significantly higher in group B than that in group A. No adverse events related to magnesium administration were observed.

CONCLUSIONS: The study design, which treated a single chemotherapy course as a study object, failed to detect a statistically significant benefit of Mg supplementation for preventing CIN in pediatric cancer patients.

TRIAL REGISTRATION: JRCT ( https://jrct.niph.go.jp/ ) Identifier UMIN000029215 jRCTs031180251. UMIN-CTR ( http://www.umin.ac.jp/icdr/index.html ) Identifier UMIN000029215.

PMID:38564107 | DOI:10.1007/s10147-024-02489-0

3D Print Med. 2024 Apr 2;10(1):10. doi: 10.1186/s41205-024-00208-z.

ABSTRACT

BACKGROUND: Femoral head avascular necrosis (AVN), or death of femoral head tissue due to a lack of blood supply, is a leading cause of total hip replacement for non-geriatric patients. Core decompression (CD) is an effective treatment to re-establish blood flow for patients with AVN. Techniques aimed at improving its efficacy are an area of active research. We propose the use of 3D printed drill guides to accurately guide therapeutic devices for CD.

METHODS: Using femur sawbones, image processing software, and 3D modeling software, we created a custom-built device with pre-determined drill trajectories and tested the feasibility of the 3D printed drill guides for CD. A fellowship trained orthopedic surgeon used the drill guide to position an 8 ga, 230 mm long decompression device in the three synthetic femurs. CT scans were taken of the sawbones with the drill guide and decompression device. CT scans were processed in the 3D modeling software. Descriptive statistics measuring the angular and needle-tip deviation were compared to the original virtually planned model.

RESULTS: Compared to the original 3D model, the trials had a mean displacement of 1.440 ± 1.03 mm and a mean angle deviation of 1.093 ± 0.749º.

CONCLUSIONS: The drill guides were demonstrated to accurately guide the decompression device along its predetermined drill trajectory. Accuracy was assessed by comparing values to literature-reported values and considered AVN lesion size. This study demonstrates the potential use of 3D printing technology to improve the efficacy of CD techniques.

PMID:38564090 | DOI:10.1186/s41205-024-00208-z

EJNMMI Res. 2024 Apr 2;14(1):34. doi: 10.1186/s13550-024-01088-4.

ABSTRACT

BACKGROUND: In juvenile systemic lupus erythematosus (j-SLE) with neuropsychiatric (NP) symptoms, there is a lack of diagnostic biomarkers. Thus, we study whether PET-FDG may identify any metabolic dysfunction in j-NPSLE.

METHODS: A total of 19 ¹⁸FDG-PET exams were consecutively performed using PET-MRI system in 11 non-sedated patients presenting with j-NPSLE (11-18y) for less than 18 months (m) and without any significant lesion at MRI. Psychiatric symptoms were scored from 0 (none) to 3 (severe) at PET time. PET images were visually analyzed and voxel-based analyses of cerebral glucose metabolism were performed using statistical parametric mapping (spm) with an age-matched control group, at threshold set > 50 voxels using both p < 0.001 uncorrected (unc.) and p < 0.05 corrected family wise error (FWE).

RESULTS: Patients exhibited mainly psychiatric symptoms, with diffuse inflammatory j-NPSLE. First PET (n = 11) was performed at a mean of 15y of age, second/third PET (n = 7/n = 1) 6 to 19 m later. PET individual analysis detected focal bilateral anomalies in 13/19 exams visually but 19/19 using spm (unc.), mostly hypermetabolic areas (18/19). A total of 15% of hypermetabolic areas identified by spm had been missed visually. PET group analysis (n = 19) did not identify any hypometabolic area, but a large bilateral cortico-subcortical hypermetabolic pattern including, by statistical decreasing order (unc.), thalamus, subthalamic brainstem, cerebellum (vermis and cortex), basal ganglia, visual, temporal and frontal cortices. Mostly the subcortical hypermetabolism survived to FWE analysis, being most intense and extensive (51% of total volume) in thalamus and subthalamus brainstem. Hypermetabolism was strictly subcortical in the most severe NP subgroup (n = 8, scores 2-3) whereas it also extended to cerebral cortex, mostly visual, in the less severe subgroup (n = 11, scores 0-1), but difference was not significant. Longitudinal visual analysis was inconclusive due to clinical heterogeneity.

CONCLUSIONS: j-NPSLE patients showed a robust bilateral cortico-subcortical hypermetabolic network, focused subcortically, particularly in thalamus, proportionally to psychiatric features severity. Further studies with larger, but homogeneous, cohorts are needed to determine the sensitivity and specificity of this dysfunctional pattern as a potential biomarker in diffuse inflammatory j-NPSLE with normal brain MRI.

PMID:38564068 | DOI:10.1186/s13550-024-01088-4