A historical perspective of biomedical explainable AI research.

The black-box nature of most artificial intelligence (AI) models encourages the development of explainability methods to engender trust into the AI decision-making process. Such methods can be broadly categorized into two main types: post hoc explanations and inherently interpretable algorithms. We aimed at analyzing the possible associations between COVID-19 and the push of explainable AI (XAI) to the forefront of biomedical research. We automatically extracted from the PubMed database biomedical XAI studies related to concepts of causality or explainability and manually labeled 1,603 papers with respect to XAI categories. To compare the trends pre- and post-COVID-19, we fit a change point detection model and evaluated significant changes in publication rates. We show that the advent of COVID-19 in the beginning of 2020 could be the driving factor behind an increased focus concerning XAI, playing a crucial role in accelerating an already evolving trend. Finally, we present a discussion with future societal use and impact of XAI technologies and potential future directions for those who pursue fostering clinical trust with interpretable machine learning models.

Circulating biomarkers and cardiac function over 3 years after chemotherapy with anthracyclines: the ICOS-ONE trial.

AIMS: A multicentre trial, ICOS-ONE, showed increases above the upper limit of normality of cardiac troponin (cTn) in 27% of patients within 12 months after the end of cancer chemotherapy (CT) with anthracyclines, whether cardiac protection with enalapril was started at study entry in all (prevention arm) or only upon first occurrence on supra-normal cTn (troponin-triggered arm). The aims of the present post hoc analysis were (i) to assess whether anthracycline-based treatment could induce cardiotoxicity over 36 month follow-up and (ii) to describe the time course of three cardiovascular biomarkers (i.e. troponin I cTnI-Ultra, B-type natriuretic peptide BNP, and pentraxin 3 PTX3) and of left ventricular (LV) function up to 36 months. METHODS AND RESULTS: Eligible patients were those prescribed first-in-life CT, without evidence of cardiovascular disease, normal cTn, LV ejection fraction (EF) >50%, not on renin-angiotensin aldosterone system antagonists. Patients underwent echocardiography and blood sampling at 24 and 36 months. No differences were observed in biomarker concentration between the two study arms, 'prevention' vs. 'troponin-triggered'. During additional follow-up 13 more deaths occurred, leading to a total of 23 (9.5%), all due to a non-cardiovascular cause. No new occurrences of LV-dysfunction were reported. Two additional patients were admitted to the hospital for cardiovascular causes, both for acute pulmonary embolism. No first onset of raised cTnI-Ultra was reported in the extended follow-up. BNP remained within normal range: at 36 months was 23.4 ng/L, higher (N.S.) than at baseline, 17.6 ng/L. PTX3 peaked at 5.2 ng/mL 1 month after CT and returned to baseline values thereafter. cTnI-Ultra peaked at 26 ng/L 1 month after CT and returned to 3 ng/L until the last measurement at 36 months. All echocardiographic variables remained stable during follow-up with a median LVEF of 63% and left atrial volume index about 24 mL/m2 . CONCLUSIONS: First-in-life CT with median cumulative dose of anthracyclines of 180 mg/m2 does not seem to cause clinically significant cardiac injury, as assessed by circulating biomarkers and echocardiography, in patients aged 51 years (median), without pre-existing cardiac disease. This may suggest either a 100% efficacy of enalapril (given as preventive or troponin-triggered) or a reassuringly low absolute cardiovascular risk in this cohort of patients, which may not require intensive cardiologic follow-up.

Early palliative care and quality of life of advanced cancer patients-a multicenter randomized clinical trial.

BACKGROUND: To compare quality of life (QoL) of patients receiving early palliative care (EPC) vs. standard oncologic care (SOC). METHODS: Pragmatic, multicenter, randomized trial at five University and Community Hospital Cancer Centers in Northern Italy. Advanced non-small cell lung, gastric, pancreatic and biliary tract cancer patients diagnosed within the previous 8 weeks. In the EPC arm, visits were performed systematically by a dedicated physician/nurse palliative care (PC) team, who assessed physical and psychosocial symptoms, and enacted the necessary services. In the SOC arm, PC visits were only carried out if requested. The primary outcome was the difference in the change of QoL [Functional Assessment of Cancer Therapy-General measure (FACT-G)] from baseline to 12 weeks in the two groups. RESULTS: From November 2014 to March 2016, 281 patients were enrolled (142 EPC, 139 SOC); 218 completed FACT-G at 12 weeks. Baseline demographic and clinical characteristics were similar for the two groups. Values of FACT-G at baseline and 12 weeks were 72.3 (SD 12.6) and 70.1 (SD 15.5) for patients enrolled in the EPC arm, vs. 71.7 (SD 14.7) and 69.6 (SD 15.5) for the SOC arm, but the change scores did not differ significantly between groups. In the multivariable analysis, adjusting for QoL at baseline, two potential prospective prognostic factors were statistically significant: lung cancer (P=0.03) and interaction of living without a partner and intervention arm (P=0.01). Dying within 6 months (P<0.001) was also statistically significant. CONCLUSIONS: In this study, EPC did not improve QoL in advanced cancer patients, but our findings highlight aspects which may guide future research on EPC.

Anthracycline-induced cardiotoxicity: A multicenter randomised trial comparing two strategies for guiding prevention with enalapril: The International CardioOncology Society-one trial.

BACKGROUND: Troponin changes over time have been suggested to allow for an early diagnosis of cardiac injury ensuing cancer chemotherapy; cancer patients with troponin elevation may benefit of therapy with enalapril. It is unknown whether a preventive treatment with enalapril may further increase the benefit. METHODS: The International CardioOncology Society-one trial (ICOS-ONE) was a controlled, open-label trial conducted in 21 Italian hospitals. Patients were randomly assigned to two strategies: enalapril in all patients started before chemotherapy (CT; 'prevention' arm), and enalapril started only in patients with an increase in troponin during or after CT ('troponin-triggered' arm). Troponin was assayed locally in 2596 blood samples, before and after each anthracycline-containing CT cycle and at each study visit; electrocardiogram and echocardiogram were done at baseline, and at 1, 3, 6 and 12-month follow-up. Primary outcome was the incidence of troponin elevation above the threshold. FINDINGS: Of the 273 patients, 88% were women, mean age 51 ± 12 years. The majority (76%) had breast cancer, 3% had a history of hypertension and 4% were diabetic. Epirubicin and doxorubicin were most commonly prescribed, with median cumulative doses of 360 [270-360] and 240 [240-240] mg/m2, respectively. The incidence of troponin elevation was 23% in the prevention and 26% in the troponin-triggered group (p = 0.50). Three patients (1.1%) -two in the prevention, one in the troponin-triggered group-developed cardiotoxicity, defined as 10% point reduction of LV ejection fraction, with values lower than 50%. INTERPRETATION: Low cumulative doses of anthracyclines in adult patients with low cardiovascular risk can raise troponins, without differences between the two strategies of giving enalapril. Considering a benefit of enalapril in the prevention of LV dysfunction, a troponin-triggered strategy may be more convenient.