BACKGROUND: Liver histopathologic assessment is the accepted surrogate endpoint in NASH trials; however, the scoring of NASH Clinical Research Network (CRN) histologic parameters is limited by intraobserver and interobserver variability. We designed a consensus panel approach to minimize variability when using this scoring system. We assessed agreement between readers, estimated linear weighted kappas between 2 panels, compared them with published pairwise kappa estimates, and addressed how agreement or disagreement might impact the precision and validity of the surrogate efficacy endpoint in NASH trials. METHODS: Two panels, each comprising 3 liver fellowship-trained pathologists who underwent NASH histology training, independently evaluated scanned whole slide images, scoring fibrosis, inflammation, hepatocyte ballooning, and steatosis from baseline and month 18 biopsies for 100 patients from the precirrhotic NASH study REGENERATE. The consensus score for each parameter was defined as agreement by ≥2 pathologists. If consensus was not reached, all 3 pathologists read the slide jointly to achieve a consensus score. RESULTS: Between the 2 panels, the consensus was 97%-99% for steatosis, 91%-93% for fibrosis, 88%-92% for hepatocyte ballooning, and 84%-91% for inflammation. Linear weighted kappa scores between panels were similar to published NASH CRN values. CONCLUSIONS: A panel of 3 trained pathologists independently scoring 4 NASH CRN histology parameters produced high consensus rates. Interpanel kappa values were comparable to NASH CRN metrics, supporting the accuracy and reproducibility of this method. The high concordance for fibrosis scoring was reassuring, as fibrosis is predictive of liver-specific outcomes and all-cause mortality.

Background: Hypoglycemia rates usually increase when insulin treatment is intensified to improve glycemic control. We evaluated (post hoc) hypoglycemic rates in adult patients with type 1 diabetes (T1D) on sotagliflozin (a dual sodium-glucose cotransporter [SGLT] 1 and 2 inhibitor) in two phase 3, 52-week clinical trials (inTandem 1 and 2; NCT02384941 and NCT02421510). Materials and Methods: We analyzed rates of documented hypoglycemia (level 1, blood glucose ≥54 to <70 mg/dL) and clinically important hypoglycemia (level 2, glucose <54 mg/dL) in a patient-level pooled analysis (n = 1362) using a negative binomial model adjusted for hemoglobin A1c (HbA1c) at 52 weeks in patients receiving placebo, sotagliflozin 200 mg, and sotagliflozin 400 mg. Results: Rates of level 1 hypoglycemia events per patient-year were 58.25 (95% confidence interval: 50.26-67.50) with placebo, 44.86 (38.83-51.82; P = 0.0138 vs. placebo) with sotagliflozin 200 mg, and 45.68 (39.52-52.81; P = 0.0220) with sotagliflozin 400 mg. Sotagliflozin was also associated with lower rates of level 2 hypoglycemia: 15.95 (14.37-17.70), 11.51 (10.39-12.76; P < 0.0001), and 11.13 (10.03-12.35; P < 0.0001) for placebo and sotagliflozin 200 and 400 mg, respectively. The difference in rates of hypoglycemia with sotagliflozin versus placebo became more pronounced as HbA1c decreased. Conclusions: At week 52, level 1 and 2 hypoglycemia events were 22% to 30% less frequent with sotagliflozin added to optimized insulin therapy versus placebo in adults with T1D at any HbA1c level, with greater differences at lower HbA1c values. These findings support the use of sotagliflozin as an insulin adjunct in T1D.