Introduction: The availability of easily accessible continuous glucose monitoring (CGM) metrics can improve glycemic control in diabetes, and they may even become a viable alternative to hemoglobin A1c (HbA1c) laboratory tests in the next years. The REALISM-T1D study (REAl-Life glucoSe Monitoring in Type 1 Diabetes) was aimed at contributing, with real-world data, to a deeper understanding of these metrics, including the time in range (TIR)–HbA1c relationship, to facilitate their adoption by diabetologists in everyday practice. Research design and methods: 70 adults affected by type 1 diabetes were monitored for 1 year by means of either flash (FGM) or real-time (rtCGM) glucose monitoring devices. Follow-up visits were performed after 90, 180 and 365 days from baseline and percentage TIR^70–180 evaluated for the 90-day time period preceding each visit. HbA1c tests were also carried out in the same occasions and measured values paired with the corresponding TIR data. Results: A monovariate linear regression analysis confirms a strong correlation between TIR and HbA1c as found in previous studies, but leveraging more homogeneous data (n=146) collected in real-life conditions. Differences were determined between FGM and rtCGM devices in Pearson’s correlation (r_FGM=0.703, r_rtCGM=0.739), slope (β_1,FGM=−11.77, β_1,rtCGM=−10.74) and intercept (β_0,FGM=141.19, β_0,rtCGM=140.77) coefficients. Normality of residuals and homoscedasticity were successfully verified in both cases. Conclusions: Regression lines for two patient groups monitored through FGM and rtCGM devices, respectively, while confirming a linear relationship between TIR and A1c hemoglobin (A1C) in good accordance with previous studies, also show a statistically significant difference in the regression intercept, thus suggesting the need for different models tailored to device characteristics. The predictive power of A1C as a TIR estimator also deserves further investigations.

To overcome the performance limitations of Controller Area Network (CAN), the CAN with Flexible Data-rate (CAN FD) specification has been recently released, which features increased network throughput by combining bit rate overclocking and frame oversizing. CAN FD ensures good backward compatibility with existing applications and higher protocol layers conceived for CAN. Unfortunately, high-performance transmission services shall not be used in networks where also legacy CAN controllers are connected, as doing so raises serious compatibility problems, which in turn undermine system reliability severely and lead to unsafe behavior. In this paper, the above problem is analyzed carefully and a number of solutions are proposed, which carry out a negotiation among CAN nodes so as to improve interoperability between legacy and CAN FD controllers.

Cyber-Physical Systems are usually subject to dependability requirements such as safety and reliability constraints. Over the last 50 years, a body of efficient fault-tolerance mechanisms has been devised to handle faults occurring at run-time. However, properly implementing those mechanisms is a time-consuming task that requires a great deal of know-how. In this paper, we propose a general framework which allows system designers to decouple functional and non-functional concerns, and express non-functional properties at design time using domain-specific languages. In the spirit of generative programming, functional models are then automatically "augmented" with dependability mechanisms. Importantly, the real-time behavior of the initial models in terms of sampling times and meeting deadlines is preserved. The practicality of the approach is demonstrated with the automated implementation of one prominent software fault-tolerance pattern, namely N-Version Programming, in the CPAL model-driven engineering workflow.

The protection of ICT systems is a topic whose importance has been dramatically growing in the last years. In fact, their increased interconnection, both internal and through the Internet, besides having provided higher and inexpensive flexibility (e.g. remote access, management and configuration), has also made them more subject to a wide range of (remotely exploitable) attacks and damages. This paper shows how security policies for industrial systems can be defined in the framework of standard recommendations, and especially checked with the help of an automatic analysis tool, thus leading to a set of security policies whose design and implementation match the expected security requirements. In particular, a suitable example highlights how the tool can profitably help through the steps from the design to the implementation of security policies in an industrial network.