Prediction of postoperative complications after oesophagectomy using machine-learning methods

BACKGROUND

Oesophagectomy is an operation with a high risk of postoperative complications. The aim of this single-centre retrospective study was to apply machine-learning methods to predict complications (Clavien-Dindo grade IIIa or higher) and specific adverse events.

METHODS

Patients with resectable adenocarcinoma or squamous cell carcinoma of the oesophagus and gastro-oesophageal junction who underwent Ivor Lewis oesophagectomy between 2016 and 2021 were included. The tested algorithms were logistic regression after recursive feature elimination, random forest, k-nearest neighbour, support vector machine, and neural network. The algorithms were also compared with a current risk score (the Cologne risk score).

RESULTS

457 patients had Clavien-Dindo grade IIIa or higher complications (52.9 per cent) versus 407 patients with Clavien-Dindo grade 0, I, or II complications (47.1 per cent). After 3-fold imputation and 3-fold cross-validation, the overall accuracies were: logistic regression after recursive feature elimination, 0.528; random forest, 0.535; k-nearest neighbour, 0.491; support vector machine, 0.511; neural network, 0.688; and Cologne risk score, 0.510. For medical complications, the results were: logistic regression after recursive feature elimination, 0.688; random forest, 0.664; k-nearest neighbour, 0.673; support vector machine, 0.681; neural network, 0.692; and Cologne risk score, 0.650. For surgical complications, the results were: logistic regression after recursive feature elimination, 0.621; random forest, 0.617; k-nearest neighbour, 0.620; support vector machine, 0.634; neural network, 0.667; and Cologne risk score, 0.624. The calculated area under the curve of the neural network was 0.672 for Clavien-Dindo grade IIIa or higher, 0.695 for medical complications, and 0.653 for surgical complications.

CONCLUSION

The neural network scored the highest accuracies compared with all of the other models for the prediction of postoperative complications after oesophagectomy.

Circadian transcription in liver

Circadian rhythms regulate a wide range of cellular, physiological, metabolic and behavioral activities in mammals. The complexity of tissue- and day-time specific regulation of thousands of clock controlled genes (CCGs) suggests that many transcriptional regulators are involved. Our bioinformatic analysis is based on two published DNA-array studies from mouse liver. We search overrepresented transcription factor binding sites in promoter regions of CCGs using GC-matched controls. Analyzing a large set of CCG promoters, we find known motifs such as E-boxes, D-boxes and cAMP responsive elements. In addition, we find overrepresented GC-rich motifs (Sp1, ETF, Nrf1), AT-rich motifs (TBP, Fox04, MEF-2), Y-box motifs (NF-Y, C/EBP) and cell cycle regulators (E2F, Elk-1). In a subset of system-driven genes, we find overrepresented motifs of the serum response factor SRF and the estrogen receptor ER. The analysis of published ChIP data reveals that some of our predicted regulators (C/EBP, E2F, HNF-1, Myc, MEF-2) target relatively many clock controlled genes. Our analysis of CCG promoters contributes to an understanding of the complex transcriptional regulation of circadian rhythms in liver.

Macrophages, microglia, and astrocytes are rapidly activated after crush injury of the goldfish optic nerve: a light and electron microscopic analysis

Several matrix and adhesion molecules in fish optic nerve, which are constitutively expressed, are increased during axonal regeneration and are primarily associated with nonneuronal cells (W.P. Battisti, Y. Shinar, M. Schwartz, P. Levitt, and M. Murray [1992] J. Neurocytol. 21:557-573). The current study examines the reactions of specific cell types to optic nerve crush and axonal regeneration. The goldfish optic nerve contains macroglia and microglia as well as a population of monocyte-derived cells (granular macrophages) unique to goldfish. Two cell types were OX-42 positive (granular macrophages and microglia), indicating monocyte lineage, each with a distinct morphology and distribution within the nerve. Within hours of the optic nerve crush, the number of OX-42-labeled cell profiles increased near the crush site, remained elevated during the time axons were elongating, and then declined. Microglia, but not granular macrophages, were phagocytically active. Astrocytes are readily identified in the normal optic nerve, but they exhibited marked morphologic changes within hours of injury, which is consistent with the contribution these cells make to the altered environment. Oligodendroglia could not be reliably identified in regenerating optic nerves until myelin was formed. A comparison of the distribution of OX-42-labeled cells with that of transforming growth factor beta-1 (TGF-beta 1) and tenascin suggests that these molecules are expressed by granular macrophages. Tenascin staining may be additionally associated with astrocytes and/or microglia. The rapid response of these nonneuronal cells to injury, their rapid phagocytic activity, and the secretion of growth-promoting factors by these cells likely contributes to the environment that supports robust regeneration by optic axons in the goldfish.

Transient and permanent patterns of expression of the low-affinity neurotrophin receptor in the interpeduncular nucleus of the rat

There is a prominent cholinergic projection from the medial habenular to the interpeduncular nucleus (IPN) which develops postnatally. In this study we examined the developmental course of expression of the low-affinity neurotrophin receptor (LANR), a receptor that binds to all members of the neurotrophin family, in the IPN. Three systems express LANR in the IPN. The cholinergic habenular axons that project to the intermediate and central subnuclei of the IPN are immunoreactive only during the time that the axons are growing and forming characteristic crest synapses in the intermediate subnuclei. The dorsomedial (DM) subnuclei, which are neither cholinergic targets nor contain cholinergic neurons, develop LANR immunoreactivity postnatally and the expression remains high in the adult. The walls of the prominent system of arterioles and venules that penetrate the IPN and ascend through the intermediate subnuclei are strongly immunoreactivity during the time of active angiogenesis and retain detectable but diminished levels of immunoreactivity in the adult. The LANR immunoreactivity seen in the vessels is likely to be associated with the peripheral sympathetic axons that innervate the smooth muscle in the vessels. These three systems within the same nucleus, which differ in phenotype, develop neurotrophin receptor immunoreactivity contemporaneously but their levels of expression differ in the adult, suggesting that they are regulated in different ways, possibly by different members of the neurotrophin family.