ABSTRACT
The objective of the present study was to assess the levels of circulating cytokines in patients with diffuse large B-cell lymphoma (DLBCL), and to examine the associations between the cytokine levels, clinicopathological manifestations and patient prognosis. The study enrolled 49 patients with DLBCL, 11 patients with chronic lymphocytic leukemia/small lymphocytic lymphoma and 67 healthy controls from Zhejiang Provincial People's Hospital (Hangzhou, China) between January 2017 and January 2020. The serum levels of interleukin (IL)-2, IL-4, IL-6, IL-10, IL-17, tumor necrosis factor (TNF)-α and interferon (IFN)-γ were measured using flow cytometry. The IL-6, IL-10 and IFN-γ levels were significantly raised in patients with DLBCL compared with those in the healthy controls (P<0.05). The levels of IL-10 were significantly higher in patients with raised levels of circulating lactate dehydrogenase (P<0.05), while increases in both IL-6 and IL-10 were associated with raised C-reactive protein (CRP) levels, with IL-6 levels positively associated with those of serum CRP (P<0.01; r=0.66). Additionally, International Prognostic Index (IPI) risk stratification of patients with DLBCL was strongly associated with circulating IL-6 and IL-10 levels. Raised IL-6, IL-10 and TNF-α levels were linked with worse short-term treatment efficacies (P<0.05). Moreover, the accuracy of the model predicting short-term treatment response in patients with DLBCL, obtained using the support vector machine algorithm, was 81.63%. It was also found that raised serum IL-6 and IL-10 levels, together with reduced levels of IL-17, were associated with survival of <1 year in patients with DLBCL (P<0.05), although no significant link was found between cytokine levels and long-term overall survival. In conclusion, the serum levels of IL-6, IL-10, IL-17, TNF-α and IFN-γ can potentially serve as biological indicators of DLBCL tumor immune status, and combined application with the IPI score can be a robust prognostic indicator in patients with DLBCL.
ABSTRACT
Summarization models often generate text that is poorly calibrated to quality metrics because they are trained to maximize the likelihood of a single reference (MLE). To address this, recent work has added a calibration step, which exposes a model to its own ranked outputs to improve relevance or, in a separate line of work, contrasts positive and negative sets to improve faithfulness. While effective, much of this work has focused on how to generate and optimize these sets. Less is known about why one setup is more effective than another. In this work, we uncover the underlying characteristics of effective sets. For each training instance, we form a large, diverse pool of candidates and systematically vary the subsets used for calibration fine-tuning. Each selection strategy targets distinct aspects of the sets, such as lexical diversity or the size of the gap between positive and negatives. On three diverse scientific long-form summarization datasets (spanning biomedical, clinical, and chemical domains), we find, among others, that faithfulness calibration is optimal when the negative sets are extractive and more likely to be generated, whereas for relevance calibration, the metric margin between candidates should be maximized and surprise-the disagreement between model and metric defined candidate rankings-minimized. Code to create, select, and optimize calibration sets is available at https://github.com/griff4692/calibrating-summaries.