Limitations of GPT-3.5 and GPT-4 in Applying Fleischner Society Guidelines to Incidental Lung Nodules.

Purpose: To evaluate the accuracy of GPT-3.5, GPT-4, and a fine-tuned GPT-3.5 model in applying Fleischner Society recommendations to lung nodules. Methods: We generated 10 lung nodule descriptions for each of the 12 nodule categories from the Fleischner Society guidelines, incorporating them into a single fictitious report (n = 120). GPT-3.5 and GPT-4 were prompted to make follow-up recommendations based on the reports. We then incorporated the full guidelines into the prompts and re-submitted them. Finally, we re-submitted the prompts to a fine-tuned GPT-3.5 model. Results were analyzed using binary accuracy analysis in R. Results: GPT-3.5 accuracy in applying Fleischner Society guidelines was 0.058 (95% CI: 0.02, 0.12). GPT-4 accuracy was improved at 0.15 (95% CI: 0.09, 0.23; P = .02 for accuracy comparison). In recommending PET-CT and/or biopsy, both GPT-3.5 and GPT-4 had an F-score of 0.00. After explicitly including the Fleischner Society guidelines in the prompt, GPT-3.5 and GPT-4 significantly improved their accuracy to 0.42 (95% CI: 0.33, 0.51; P < .001) and to 0.66 (95% CI: 0.57, 0.74; P < .001), respectively. GPT-4 remained significantly better than GPT-3.5 (P < .001). The fine-tuned GPT-3.5 model accuracy was 0.46 (95% CI: 0.37, 0.55), not different from the GPT-3.5 model with guidelines included (P = .53). Conclusion: GPT-3.5 and GPT-4 performed poorly in applying widely known guidelines and never correctly recommended biopsy. Flawed knowledge and reasoning both contributed to their poor performance. While GPT-4 was more accurate than GPT-3.5, its inaccuracy rate was unacceptable for clinical practice. These results underscore the limitations of large language models for knowledge and reasoning-based tasks.

Access-to-Care and Conscience: Conflicting or Coherent?

"Intervention" is not synonymous with "care." For an intervention to constitute care-which patients should have a right to access-it must be technically feasible and licit. Now these criteria do not prove sufficient; numerous archaic interventions remain feasible and legally permissible, yet are now bywords for spurious care. Therefore, we propound another necessary condition for an intervention to become care: the physician must rationally judge the intervention to be conducive to the patient's good. Consequently, the right of access-to-care relies on physicians being free to practice medicine in accord with their consciences, conscience being the rational faculty with which they judge the reasonableness of even mundane medical decisions. Since physicians operate as part of a community, it is further necessary to consider when central bodies may reasonably compel physicians to engage in interventions that the physician believes are not consistent with the patient's good and/or are not congruent with the purposes of medicine.

To die, to sleep, perchance to dream? A response to DeMichelis, Shaul and Rapoport.

In developing their policy on paediatric medical assistance in dying (MAID), DeMichelis, Shaul and Rapoport decide to treat euthanasia and physician-assisted suicide as ethically and practically equivalent to other end-of-life interventions, particularly palliative sedation and withdrawal of care (WOC). We highlight several flaws in the authors' reasoning. Their argument depends on too cursory a dismissal of intention, which remains fundamental to medical ethics and law. Furthermore, they have not fairly presented the ethical analyses justifying other end-of-life decisions, analyses and decisions that were generally accepted long before MAID was legal or considered ethical. Forgetting or misunderstanding the analyses would naturally lead one to think MAID and other end-of-life decisions are morally equivalent. Yet as we recall these well-developed analyses, it becomes clear that approving of some forms of sedation and WOC does not commit one to MAID. Paediatric patients and their families can rationally and coherently reject MAID while choosing palliative care and WOC. Finally, the authors do not substantiate their claim that MAID is like palliative care in that it alleviates suffering. It is thus unreasonable to use this supposition as a warrant for their proposed policy.

Estradiol Levels Are Altered in Human Immunodeficiency Virus-Infected Pregnant Women Randomized to Efavirenz-Versus Lopinavir/Ritonavir-Based Antiretroviral Therapy.

Background: Combination antiretroviral therapy (cART) use in pregnancy has been associated with hormonal dysregulation. We performed a secondary retrospective analysis of longitudinal progesterone and estradiol levels in pregnancy using specimens from the Protease Inhibitors to Reduce Malaria Morbidity in HIV-infected Pregnant Women study, which randomized Ugandan human immunodeficiency virus (HIV)-infected ART-naive women to initiate either lopinavir/ritonavir (LPV/r)-based or efavirenz (EFV)-based cART. Methods: Three hundred twenty-six women (160 randomized to the EFV arm and 166 women to the LPV/r arm) with at least 1 plasma sample collected during pregnancy were included. Enrollment samples collected prior to cART initiation were used as a cART-naive comparator group. Hormone levels were quantified by enzyme-linked immunosorbent assay. Results: Estradiol levels were differentially affected by the 2 cART regimens. Exposure to LPV/r was associated with an increase in estradiol (P < .0001), whereas exposure to EFV was associated with a decrease in estradiol (P < .0001), relative to the cART-naive gestationally matched comparator group. Lower estradiol levels correlated with small for gestational age (SGA) (P = .0019) and low birth weight (P = .019) in the EFV arm, while higher estradiol levels correlated with SGA in the LPV/r arm (P = .027). Although progesterone levels were similar between treatment arms, we observed an association between SGA and lower progesterone in the LPV/r arm (P = .04). No association was observed between hormone levels and preterm birth in either arm. Levels of progesterone and estradiol were lower in cases of stillbirth, and levels of both hormones declined immediately prior to stillbirth in 5 of 8 cases. Conclusions: Combination ART regimens differentially affect estradiol levels in pregnancy, a hormone critical to the maintenance of a healthy pregnancy. Identifying cART regimens that minimize perinatal HIV transmission without contributing to hormonal dysregulation represents an urgent public health priority. Clinical Trials Registration: NCT00993031.

Altered angiogenesis as a common mechanism underlying preterm birth, small for gestational age, and stillbirth in women living with HIV.

BACKGROUND: Angiogenic processes in the placenta are critical regulators of fetal growth and impact birth outcomes, but there are limited data documenting these processes in HIV-infected women or women from low-resource settings. OBJECTIVE: We sought to determine whether angiogenic factors are associated with adverse birth outcomes in HIV-infected pregnant women started on antiretroviral therapy. STUDY DESIGN: This is a secondary analysis of samples collected as part of a clinical trial randomizing pregnant women and adolescents infected with HIV to lopinavir/ritonavir-based (n = 166) or efavirenz-based (n = 160) antiretroviral therapy in Tororo, Uganda. Pregnant women living with HIV were enrolled between 12-28 weeks of gestation. Plasma samples were evaluated for angiogenic biomarkers (angiopoietin-1, angiopoietin-2, vascular endothelial growth factor, soluble fms-like tyrosine kinase-1, placental growth factor, and soluble endoglin) by enzyme-linked immunosorbent assay between: 16-<20, 20-<24, 24-<28, 28-<32, 32-<36, 36-<37 weeks of gestation. The primary outcome was preterm birth. RESULTS: In all, 1115 plasma samples from 326 pregnant women and adolescents were evaluated. There were no differences in angiogenic factors according to antiretroviral therapy group (P > .05 for all). The incidence of adverse birth outcomes was 16.9% for spontaneous preterm births, 25.6% for small-for-gestational-age births, and 2.8% for stillbirth. We used linear mixed effect modelling to evaluate longitudinal changes in angiogenic factor concentrations between birth outcome groups adjusting for gestational age at venipuncture, maternal age, body mass index, gravidity, and the interaction between treatment arm and gestational age. Two angiogenic factors-soluble endoglin and placental growth factor-were associated with adverse birth outcomes. Significantly higher concentrations of soluble endoglin throughout gestation were found in study participants destined to deliver preterm [likelihood ratio test, χ2(1) = 12.28, P < .0005] and in those destined to have stillbirths [χ2(1) = 5.67, P < .02]. By contrast, significantly lower concentrations of placental growth factor throughout gestation were found in those destined to have small-for-gestational-age births [χ2(1) = 7.89, P < .005] and in those destined to have stillbirths [χ2(1) = 21.59, P < .0001]. CONCLUSION: An antiangiogenic state in the second or third trimester is associated with adverse birth outcomes, including stillbirth in women and adolescents living with HIV and receiving antiretroviral therapy.

Malaria in pregnancy alters l-arginine bioavailability and placental vascular development.

Reducing adverse birth outcomes due to malaria in pregnancy (MIP) is a global health priority. However, there are few safe and effective interventions. l-Arginine is an essential amino acid in pregnancy and an immediate precursor in the biosynthesis of nitric oxide (NO), but there are limited data on the impact of MIP on NO biogenesis. We hypothesized that hypoarginemia contributes to the pathophysiology of MIP and that l-arginine supplementation would improve birth outcomes. In a prospective study of pregnant Malawian women, we show that MIP was associated with lower concentrations of l-arginine and higher concentrations of endogenous inhibitors of NO biosynthesis, asymmetric and symmetric dimethylarginine, which were associated with adverse birth outcomes. In a model of experimental MIP, l-arginine supplementation in dams improved birth outcomes (decreased stillbirth and increased birth weight) compared with controls. The mechanism of action was via normalized angiogenic pathways and enhanced placental vascular development, as visualized by placental microcomputerized tomography imaging. These data define a role for dysregulation of NO biosynthetic pathways in the pathogenesis of MIP and support the evaluation of interventions to enhance l-arginine bioavailability as strategies to improve birth outcomes.

Brain-derived Neurotrophic Factor Is Associated With Disease Severity and Clinical Outcome in Ugandan Children Admitted to Hospital With Severe Malaria.

BACKGROUND: Malaria remains a leading cause of childhood death and neurologic disability in sub-Saharan Africa. Here, we test the hypothesis that malaria-induced alterations to circulating brain-derived neurotrophic factor (BDNF) are associated with poor clinical outcomes in children with severe malaria. METHODS: We quantified BDNF (by enzyme-linked immunosorbent assay) in plasma samples collected [at presentation (day 1), day 3 and day 14], during a prospective study of Ugandan children admitted to hospital with severe malaria (n = 179). RESULTS: BDNF concentration at presentation (day 1) was lower in children with cerebral malaria (P < 0.01), coma (P < 0.01), Lambaréné Organ Dysfunction Score >1 (P < 0.05) and respiratory distress (P < 0.01). Higher BDNF concentration at presentation was associated with shorter time to coma recovery [hazard ratio = 1.655 (1.194-2.293); P = 0.002] and a reduced odds ratio of disability [0.50 (0.27-0.94); P = 0.047] and death [0.45 (0.22-0.92); P = 0.035]. BDNF concentration was lower on day 1 and increased in children surviving severe malaria (day 14; P < 0.0001). CONCLUSIONS: Our findings provide the new evidence linking circulating BDNF with disease severity, coma recovery and clinical outcome in children with severe malaria.