A paradigm shift?-On the ethics of medical large language models.

After a wave of breakthroughs in image-based medical diagnostics and risk prediction models, machine learning (ML) has turned into a normal science. However, prominent researchers are claiming that another paradigm shift in medical ML is imminent-due to most recent staggering successes of large language models-from single-purpose applications toward generalist models, driven by natural language. This article investigates the implications of this paradigm shift for the ethical debate. Focusing on issues like trust, transparency, threats of patient autonomy, responsibility issues in the collaboration of clinicians and ML models, fairness, and privacy, it will be argued that the main problems will be continuous with the current debate. However, due to functioning of large language models, the complexity of all these problems increases. In addition, the article discusses some profound challenges for the clinical evaluation of large language models and threats to the reproducibility and replicability of studies about large language models in medicine due to corporate interests.

Uncertainty, Evidence, and the Integration of Machine Learning into Medical Practice.

In light of recent advances in machine learning for medical applications, the automation of medical diagnostics is imminent. That said, before machine learning algorithms find their way into clinical practice, various problems at the epistemic level need to be overcome. In this paper, we discuss different sources of uncertainty arising for clinicians trying to evaluate the trustworthiness of algorithmic evidence when making diagnostic judgments. Thereby, we examine many of the limitations of current machine learning algorithms (with deep learning in particular) and highlight their relevance for medical diagnostics. Among the problems we inspect are the theoretical foundations of deep learning (which are not yet adequately understood), the opacity of algorithmic decisions, and the vulnerabilities of machine learning models, as well as concerns regarding the quality of medical data used to train the models. Building on this, we discuss different desiderata for an uncertainty amelioration strategy that ensures that the integration of machine learning into clinical settings proves to be medically beneficial in a meaningful way.

Enabling Fairness in Healthcare Through Machine Learning.

The use of machine learning systems for decision-support in healthcare may exacerbate health inequalities. However, recent work suggests that algorithms trained on sufficiently diverse datasets could in principle combat health inequalities. One concern about these algorithms is that their performance for patients in traditionally disadvantaged groups exceeds their performance for patients in traditionally advantaged groups. This renders the algorithmic decisions unfair relative to the standard fairness metrics in machine learning. In this paper, we defend the permissible use of affirmative algorithms; that is, algorithms trained on diverse datasets that perform better for traditionally disadvantaged groups. Whilst such algorithmic decisions may be unfair, the fairness of algorithmic decisions is not the appropriate locus of moral evaluation. What matters is the fairness of final decisions, such as diagnoses, resulting from collaboration between clinicians and algorithms. We argue that affirmative algorithms can permissibly be deployed provided the resultant final decisions are fair.

Can Robots Do Epidemiology? Machine Learning, Causal Inference, and Predicting the Outcomes of Public Health Interventions.

This paper argues that machine learning (ML) and epidemiology are on collision course over causation. The discipline of epidemiology lays great emphasis on causation, while ML research does not. Some epidemiologists have proposed imposing what amounts to a causal constraint on ML in epidemiology, requiring it either to engage in causal inference or restrict itself to mere projection. We whittle down the issues to the question of whether causal knowledge is necessary for underwriting predictions about the outcomes of public health interventions. While there is great plausibility to the idea that it is, conviction that something is impossible does not by itself motivate a constraint to forbid trying. We disambiguate the possible motivations for such a constraint into definitional, metaphysical, epistemological, and pragmatic considerations and argue that "Proceed with caution" (rather than "Stop!") is the outcome of each. We then argue that there are positive reasons to proceed, albeit cautiously. Causal inference enforces existing classification schema prior to the testing of associational claims (causal or otherwise), but associations and classification schema are more plausibly discovered (rather than tested or justified) in a back-and-forth process of gaining reflective equilibrium. ML instantiates this kind of process, we argue, and thus offers the welcome prospect of uncovering meaningful new concepts in epidemiology and public health-provided it is not causally constrained.

On Algorithmic Fairness in Medical Practice.

The application of machine-learning technologies to medical practice promises to enhance the capabilities of healthcare professionals in the assessment, diagnosis, and treatment, of medical conditions. However, there is growing concern that algorithmic bias may perpetuate or exacerbate existing health inequalities. Hence, it matters that we make precise the different respects in which algorithmic bias can arise in medicine, and also make clear the normative relevance of these different kinds of algorithmic bias for broader questions about justice and fairness in healthcare. In this paper, we provide the building blocks for an account of algorithmic bias and its normative relevance in medicine.

Randomised controlled trials in medical AI: ethical considerations.

In recent years, there has been a surge of high-profile publications on applications of artificial intelligence (AI) systems for medical diagnosis and prognosis. While AI provides various opportunities for medical practice, there is an emerging consensus that the existing studies show considerable deficits and are unable to establish the clinical benefit of AI systems. Hence, the view that the clinical benefit of AI systems needs to be studied in clinical trials-particularly randomised controlled trials (RCTs)-is gaining ground. However, an issue that has been overlooked so far in the debate is that, compared with drug RCTs, AI RCTs require methodological adjustments, which entail ethical challenges. This paper sets out to develop a systematic account of the ethics of AI RCTs by focusing on the moral principles of clinical equipoise, informed consent and fairness. This way, the objective is to animate further debate on the (research) ethics of medical AI.

How competitors become collaborators-Bridging the gap(s) between machine learning algorithms and clinicians.

For some years, we have been witnessing a steady stream of high-profile studies about machine learning (ML) algorithms achieving high diagnostic accuracy in the analysis of medical images. That said, facilitating successful collaboration between ML algorithms and clinicians proves to be a recalcitrant problem that may exacerbate ethical problems in clinical medicine. In this paper, we consider different epistemic and normative factors that may lead to algorithmic overreliance within clinical decision-making. These factors are false expectations, the miscalibration of uncertainties, non-explainability, and the socio-technical context within which the algorithms are utilized. Moreover, we identify different desiderata for bridging the gap between ML algorithms and clinicians. Further, we argue that there is an intriguing dialectic in the collaboration between clinicians and ML algorithms. While it is the algorithm that is supposed to assist the clinician in diagnostic tasks, successful collaboration will also depend on adjustments on the side of the clinician.

On the ethics of algorithmic decision-making in healthcare.

In recent years, a plethora of high-profile scientific publications has been reporting about machine learning algorithms outperforming clinicians in medical diagnosis or treatment recommendations. This has spiked interest in deploying relevant algorithms with the aim of enhancing decision-making in healthcare. In this paper, we argue that instead of straightforwardly enhancing the decision-making capabilities of clinicians and healthcare institutions, deploying machines learning algorithms entails trade-offs at the epistemic and the normative level. Whereas involving machine learning might improve the accuracy of medical diagnosis, it comes at the expense of opacity when trying to assess the reliability of given diagnosis. Drawing on literature in social epistemology and moral responsibility, we argue that the uncertainty in question potentially undermines the epistemic authority of clinicians. Furthermore, we elucidate potential pitfalls of involving machine learning in healthcare with respect to paternalism, moral responsibility and fairness. At last, we discuss how the deployment of machine learning algorithms might shift the evidentiary norms of medical diagnosis. In this regard, we hope to lay the grounds for further ethical reflection of the opportunities and pitfalls of machine learning for enhancing decision-making in healthcare.

Secondary immune thrombocytopenia (ITP) as an initial presentation of Whipple's disease.

Immune thrombocytopenia (ITP) is a heterogeneous autoimmune disease characterized by low platelet count that has been associated with a number of chronic infections but rarely described as a manifestation of Whipple's disease (WD). We present a case of Whipple's disease in a patient initially diagnosed with ITP. A 46-year old male in the fifth decade of life presented with presumed idiopathic ITP and was treated with several therapies including corticosteroids, rituximab, and thrombopoietin receptor agonists. Several years later, he developed weight loss and worsening arthralgias. He was found to have evidence of WD in a jejunal lymph node, the duodenum, and the cerebral spinal fluid (CSF). His diagnosis of WD, as a cause of secondary ITP, came a full 8 years after he was discovered to have thrombocytopenia and over 4 years after he was diagnosed with ITP. WD is an uncommon, multiorgan system disease caused by the actinomycete Tropheryma whipplei. Whipple's disease presents a diagnostic challenge due to the wide array of possible presenting clinical manifestations, as well as a prolonged time course with separation of symptoms over many years. While T. whipplei is ubiquitous in the environment, few individuals develop clinical disease, raising the prospect that select immunodeficiencies, both singular or in combination, may play a role in infection. While rare, in the appropriate clinical setting, one should consider infection with T. whipplei in addition to other chronic infections as a cause of secondary ITP regardless of how long ago the diagnosis of ITP was made.

Targeting Lymph Node Retrieval and Assessment in Stage II Colon Cancer: A Quality Outcome Community-Based Cancer Center Study.

PURPOSE: Adequate lymph node evaluation is required for the proper staging of colon cancer. The current recommended number of lymph nodes that should be retrieved and assessed is 12. METHODS: The multidisciplinary Gastrointestinal Tumor Board at the Derrick L. Davis Forsyth Regional Cancer Center reviewed and recommended that a minimum of 12 lymph nodes be examined in all cases of colon cancer to ensure proper staging. This recommendation occurred at the end of the first quarter of 2005. To ensure this new standard was being followed, an outcomes study looking at the number of lymph nodes evaluated in stage II colon cancer was initiated. All patients with stage II colon cancer diagnosed between 2004 and 2006 were reviewed. RESULTS: There was a statistically significant improvement in the number of stage II colon cancer patients with 12 or more lymph nodes evaluated. Before the Gastrointestinal Tumor Board's recommendation, 49% (40 out of 82 patients) had 12 or more lymph nodes sampled. The median number of lymph nodes evaluated was 11. After the Gastrointestinal Tumor Board's recommendation, 79% (70 out of 88 patients) had 12 or more lymph nodes sampled. The median number of lymph nodes was 16. CONCLUSION: Multidisciplinary tumor boards can impact the quality of care of patients as demonstrated in this study. Although we do not yet have survival data on these patients, based on the previous literature referenced in this article, we would expect to see an improvement in survival rates in patients with 12 or more nodes retrieved and assessed.

Infusion of palonosetron plus dexamethasone for the prevention of chemotherapy-induced nausea and vomiting.

Serotonin (5-HT3) receptor antagonists are the foundation of standard antiemetic care for cancer patients receiving emetogenic chemotherapy. To enhance the efficacy of these supportive care agents, dexamethasone is routinely admixed with the 5-HT3 receptor antagonist, which is administered by intravenous infusion before chemotherapy begins. This phase II study evaluated the safety and efficacy of intravenous palonosetron admixed with dexamethasone to prevent chemotherapy-induced nausea and vomiting (CINV) in patients receiving moderately emetogenic chemotherapy. Cancer patients received palonosetron 0.25 mg plus dexamethasone 8 mg admixed in 50 mL of infusion solution before receiving at least one qualifying chemotherapeutic agent (cyclophosphamide < or = 1,500 mg/m2, doxorubicin > or = 20 mg/m2, carboplatin, or oxaliplatin). Patients used diaries to record nausea and emesis experienced and rescue medications used. Of 32 participants, 27 (84%) had a complete response (no emesis and no rescue medication) during the acute (0-24 hours) interval posttherapy, 19 (59%) had a complete response during the delayed (> 24-120 hours) posttherapeutic interval, and 19 (59%) had a complete response during the overall (0-120 hours) posttreatment interval. A total of 23 patients (72%) had no emetic episodes, 16 (50%) had no nausea, and 21 (66%) used no rescue medication throughout the overall 5-day interval. The combination was well tolerated. Palonosetron plus dexamethasone given as a pretreatment infusion is effective and safe in preventing acute and delayed CINV in patients receiving moderately emetogenic chemotherapy.

Combination therapy for chemotherapy-induced nausea and vomiting in patients receiving moderately emetogenic chemotherapy: palonosetron, dexamethasone, and aprepitant.

The objective of this multicenter, phase II, open-label study was to evaluate the safety and efficacy of the newest 5-hydroxytryptamine3 (5-HT3) receptor antagonist, palonosetron, plus dexamethasone and aprepitant in preventing nausea and vomiting in patients receiving moderately emetogenic chemotherapy. Eligible patients received a single intravenous dose of palonosetron (0.25 mg on day 1 of chemotherapy), along with 3 daily oral doses of aprepitant (125 mg on day 1,80 mg on days 2 and 3) and dexamethasone (12 mg on day 1,8 mg on days 2 and 3). Efficacy and safety data were obtained from patient diaries and adverse event reporting. Fifty-eight patients were evaluable; 47% were women with breast cancer and 52% received cyclophosphamide-based chemotherapy. The proportion of patients with complete response (no emesis and no rescue medication) was 88% during the acute (0-24 hours) interval, 78% during the delayed (> 24-120 hours) interval, and 78% during the overall (0-120 hours post chemotherapy) interval. More than 90% of patients during all time intervals had no emetic episodes, and between 57% and 71% of patients reported no nausea during each of the 5 days post chemotherapy. Treatment was well tolerated, with no unexpected adverse events. These data demonstrate that palonosetron in combination with dexamethasone and aprepitant is safe and highly effective in preventing chemotherapy-induced nausea and vomiting in the days following administration of moderately emetogenic chemotherapy.

Postoperative upper airway obstruction after recovery of the train of four ratio of the adductor pollicis muscle from neuromuscular blockade.

Anesthetics, and even minimal residual neuromuscular blockade, may lead to upper airway obstruction (UAO). In this study we assessed by spirometry in patients with a train-of-four (TOF) ratio >0.9 the incidence of UAO (i.e., the ratio of maximal expiratory flow and maximal inspiratory flow at 50% of vital capacity [MEF50/MIF50] >1) and determined if UAO is induced by neuromuscular blockade (defined by a forced vital capacity [FVC] fade, i.e., a decrease in values of FVC from the first to the second consecutive spirometric maneuver of > or =10%). Patients received propofol and opioids for anesthesia. Spirometry was performed by a series of 3 repetitive spirometric maneuvers: the first before induction (under midazolam premedication), the second after tracheal extubation (TOF ratio: 0.9 or more), and the third 30 min later. Immediately after tracheal extubation and 30 min later, 48 and 6 of 130 patients, respectively, were not able to perform spirometry appropriately because of sedation. The incidence of UAO increased significantly (P < 0.01) from 82 of 130 patients (63%) at preinduction baseline to 70 of 82 patients (85%) after extubation, and subsequently decreased within 30 min to values observed at baseline (80 of 124 patients, 65%). The mean maximal expiratory flow and maximal inspiratory flow at 50% of vital capacity ratio after tracheal extubation was significantly increased from baseline (by 20%; 1.39 +/- 1.01 versus 1.73 +/- 1.02; P < 0.01), and subsequently decreased significantly to values observed at baseline (1.49 +/- 0.93). A statistically significant FVC fade was not present, and a FVC fade of > or =10% was observed in only 2 patients after extubation. Thus, recovery of the TOF ratio to 0.9 predicts with high probability an absence of neuromuscular blocking drug-induced UAO, but outliers, i.e., persistent effects of neuromuscular blockade on upper airway integrity despite recovery of the TOF ratio, may still occur.

Efficacy and safety analysis of epoetin alfa in patients with small-cell lung cancer: a randomized, double-blind, placebo-controlled trial.

PURPOSE: This randomized, double-blind, placebo-controlled trial (N93-004) evaluated the effects of epoetin alfa on tumor response to chemotherapy and survival in patients with small-cell lung cancer (SCLC). PATIENTS AND METHODS: Adult patients with hemoglobin < or = 14.5 g/dL starting chemotherapy received epoetin alfa 150 U/kg or placebo subcutaneously 3 times weekly until 3 weeks after completion of chemotherapy. Survival was assessed for 3 years. The primary end point was the proportion of patients with complete or partial response after three chemotherapy cycles. RESULTS: The trial was terminated prematurely after 224 of a projected 400 patients were accrued. Baseline characteristics were similar between groups. Epoetin alfa and placebo patients (n = 109 and n = 115, respectively) had mean baseline hemoglobin of 12.8 g/dL and 13.0 g/dL, respectively. Overall tumor response was similar between the epoetin alfa and placebo groups after three chemotherapy cycles (72% and 67%, respectively; 95% CI of difference, -6% to 18%) and after completion of chemotherapy (60% and 56%, respectively; 95% CI of difference, -9% to 17%). Epoetin alfa and placebo groups had similar median overall survival (10.5 and 10.4 months, respectively) and overall mortality (91.7% and 87.8%, respectively; hazard ratio, 1.172; 95% CI, 0.887 to 1.549; P = .264). Hemoglobin was maintained in the prechemotherapy range in epoetin alfa patients, but decreased substantially in placebo patients. Fewer epoetin alfa patients than placebo patients required transfusion. CONCLUSION: These results suggest that in newly diagnosed patients with SCLC epoetin alfa does not affect tumor response to chemotherapy or survival. However, the early trial closure makes these conclusions preliminary.