A risk analysis of alpelisib (ALP)-induced hyperglycemia (HG) using baseline factors in patients (pts) with advanced solid tumours and breast cancer (BC): A pooled analysis of X2101 and SOLAR-1

ALP + fulvestrant is approved for PIK3CA-mutated, HR+/HER2– advanced BC in the US and EU. HG is an on-target adverse event (AE) of ALP and the most common all grade (G) and G3/4 AE in the phase III SOLAR-1 trial. Antihyperglycemic agents (AHAs) and ALP dose modifications (mods) and discontinuations (discs) were used to manage this AE (Table). We present a pooled risk factor model using safety sets of the phase I X2101 (NCT01219699) and SOLAR-1 (NCT02437318) trials and its application to SOLAR-1 data. Pts (505) were randomly divided into training (405 pts) and testing sets (100 pts). Baseline factors for HG were identified, multiple models tested, and a random forest model was used to categorize pts as high or low risk for G3/4 HG based on 5 baseline factors (fasting plasma glucose [FPG], BMI, HbA1c, monocyte counts, and age). Performance metrics will be presented. The training set identified G3/4 HG in 126/131 high (96.2%) vs 3/274 low risk (1.09%) pts, and was validated in the testing set with 20/33 high (60.6%) vs 12/67 low risk (17.9%) pts. A 2-month (mo) analysis of G3/4 HG risk confirmed decreased probability at lower risk scores (area under the curve: training set = 0.991, testing set = 0.767). In all pts in the ALP arm of SOLAR-1, the model found a higher incidence of all-G and G3/4 HG, use of multiple AHAs, and more dose mods and discs in the high vs low risk pts (Table). There was no difference in median PFS in high (11.0 mo) vs low risk (10.9 mo) pts in the ALP arm with PIK3CA mutations.Table: 96MOHG incidence and management, n (%)All Pts N = 284High Risk N = 106Low Risk N = 178All-G HG187 (65.8)101 (95.3)86 (48.3)G3/4 HG108 (38.0)96 (90.6)12 (6.7)Patients who received AHAs163 (57.4)94 (88.7)70 (39.3)1 AHA67 (41.1)24 (25.5)43 (61.4)2 AHA49 (30.1)31 (33.0)18 (25.7)≥3 AHA47 (28.8)39 (41.5)9 (12.9)Dose mods for any reason (reductions and/or interruption of ALP)213 (75.0)92 (86.8)122 (68.5)Permanent discs for any reason244 (85.9)90 (84.9)154 (86.5)Discs due to HG19 (6.7)15 (14.2)4 (2.2) Open table in a new tab Risk modeling identified 5 baseline factors (FPG, BMI, HbA1c, monocyte counts, and age) associated with a higher probability of ALP-induced G3/4 HG. High risk pts had higher rates of AHAs and ALP mods. This model could be used clinically to identify pts at high risk for ALP-induced HG. Regardless of risk, pts with PIK3CA mutations derived a similar PFS benefit from ALP.