Peter Pickkers

Abstract Fast and reliable detection of patients with severe and heterogeneous illnesses is a major goal of precision medicine 1,2 . Patients with leukaemia can be identified using machine learning on the basis of their blood transcriptomes 3 . However, there is an increasing divide between what is technically possible and what is allowed, because of privacy legislation 4,5 . Here, to facilitate the integration of any medical data from any data owner worldwide without violating privacy laws, we introduce Swarm Learning—a decentralized machine-learning approach that unites edge computing, blockchain-based peer-to-peer networking and coordination while maintaining confidentiality without the need for a central coordinator, thereby going beyond federated learning. To illustrate the feasibility of using Swarm Learning to develop disease classifiers using distributed data, we chose four use cases of heterogeneous diseases (COVID-19, tuberculosis, leukaemia and lung pathologies). With more than 16,400 blood transcriptomes derived from 127 clinical studies with non-uniform distributions of cases and controls and substantial study biases, as well as more than 95,000 chest X-ray images, we show that Swarm Learning classifiers outperform those developed at individual sites. In addition, Swarm Learning completely fulfils local confidentiality regulations by design. We believe that this approach will notably accelerate the introduction of precision medicine.

BACKGROUND: Selective digestive tract decontamination (SDD) and selective oropharyngeal decontamination (SOD) are infection-prevention measures used in the treatment of some patients in intensive care, but reported effects on patient outcome are conflicting. METHODS: We evaluated the effectiveness of SDD and SOD in a crossover study using cluster randomization in 13 intensive care units (ICUs), all in The Netherlands. Patients with an expected duration of intubation of more than 48 hours or an expected ICU stay of more than 72 hours were eligible. In each ICU, three regimens (SDD, SOD, and standard care) were applied in random order over the course of 6 months. Mortality at day 28 was the primary end point. SDD consisted of 4 days of intravenous cefotaxime and topical application of tobramycin, colistin, and amphotericin B in the oropharynx and stomach. SOD consisted of oropharyngeal application only of the same antibiotics. Monthly point-prevalence studies were performed to analyze antibiotic resistance. RESULTS: A total of 5939 patients were enrolled in the study, with 1990 assigned to standard care, 1904 to SOD, and 2045 to SDD; crude mortality in the groups at day 28 was 27.5%, 26.6%, and 26.9%, respectively. In a random-effects logistic-regression model with age, sex, Acute Physiology and Chronic Health Evaluation (APACHE II) score, intubation status, and medical specialty used as covariates, odds ratios for death at day 28 in the SOD and SDD groups, as compared with the standard-care group, were 0.86 (95% confidence interval [CI], 0.74 to 0.99) and 0.83 (95% CI, 0.72 to 0.97), respectively. CONCLUSIONS: In an ICU population in which the mortality rate associated with standard care was 27.5% at day 28, the rate was reduced by an estimated 3.5 percentage points with SDD and by 2.9 percentage points with SOD. (Controlled Clinical Trials number, ISRCTN35176830.)

Suppression of immune responses is necessary to limit damage to host tissue during inflammation, but it can be detrimental in specific immune responses, such as sepsis and antitumor immunity. Recently, immature myeloid cells have been implicated in the suppression of immune responses in mouse models of cancer, infectious disease, bone marrow transplantation, and autoimmune disease. Here, we report the identification of a subset of mature human neutrophils (CD11cbright/CD62Ldim/CD11bbright/CD16bright) as what we believe to be a unique circulating population of myeloid cells, capable of suppressing human T cell proliferation. These cells were observed in humans in vivo during acute systemic inflammation induced by endotoxin challenge or by severe injury. Local release of hydrogen peroxide from the neutrophils into the immunological synapse between the neutrophils and T cells mediated the suppression of T cell proliferation and required neutrophil expression of the integrin Mac-1 (αMβ2). Our data demonstrate that suppression of T cell function can be accomplished by a subset of human neutrophils that can be systemically induced in response to acute inflammation. Identification of the pivotal role of neutrophil Mac-1 and ROS in this process provides a potential target for modulating immune responses in humans.

Given that the leading clinical conditions associated with acute kidney injury (AKI), namely, sepsis, major surgery, heart failure, and hypovolemia, are all associated with shock, it is tempting to attribute all AKI to ischemia on the basis of macrohemodynamic changes. However, an increasing body of evidence has suggested that in many patients, AKI can occur in the absence of overt signs of global renal hypoperfusion. Indeed, sepsis-induced AKI can occur in the setting of normal or even increased renal blood flow. Accordingly, renal injury may not be entirely explained solely on the basis of the classic paradigm of hypoperfusion, and thus other mechanisms must come into play. Herein, we put forward a "unifying theory" to explain the interplay between inflammation and oxidative stress, microvascular dysfunction, and the adaptive response of the tubular epithelial cell to the septic insult. We propose that this response is mostly adaptive in origin, that it is driven by mitochondria, and that it ultimately results in and explains the clinical phenotype of sepsis-induced AKI.