NCOG-30. Principal component analysis and Gaussian mixture modelling identify distinct cognitive profiles in glioblastoma patients undergoing surgery
- Abstract:
- Abstract INTRODUCTION Glioblastoma patients frequently suffer from cognitive dysfunction. The effects of surgery on cognition are unclear. It is unknown whether heterogeneity in these test scores reflects underlying disease patterns and whether clinical variables can identify which patients are at risk of cognitive deterioration after surgery. METHODS Neuropsychological testing was prospectively performed in 50 patients with glioblastoma preoperatively, early postoperatively (within 1 week) and late postoperatively (at the start of chemoradiotherapy). Principal component analysis with Gaussian mixture models was used to examine the predictors of postsurgical cognitive deterioration across demographic and clinical variables. RESULTS Patients suffered from cognitive deficits across three domains: 1) attention, perception and memory (RC1), 2) attention, numerical calculation and praxis (RC2), and 3) executive function (RC3). Gaussian mixture modelling identified three cognitive profiles: 1) minimal impairment across the three domains (56%) (P1), 2) generalised impairment across all domains (32%) (P2) and 3) isolated impairments in single domains (12%) (P3). Profiles predicted postoperative deterioration across cognitive domains in mixed-effects models. P2 predicted reduced RC1 scores (β = -0.65, SE = 0.25, t = -2.54, p =.015), RC2 scores (β = -0.49, SE = 0.23, t = -2.07, p =.046) and RC3 scores (β = -0.61, SE = 0.22, t = -2.81, p =.008). P3 predicted reduced RC2 (β = -1.41, SE = 0.30, t = -4.57, p <.001) and RC3 scores (β = -1.85, SE = 0.28, t = -6.44, p <.001). Right-sided tumours predicted higher RC1 scores (β = 0.91, SE = 0.23, t = 4.00, p <.001) and RC2 scores (β = -0.70, SE = 0.20, t = -3.57, p =.001). CONCLUSIONS We identified three cognitive profiles in glioblastoma patients which predicted postoperative deterioration. Our approach offers the opportunity to stratify surgery and rehabilitation based on individualised patterns of cognitive deficits.
- Authors:
- Y Wan, R Mair, T Santarius, A Halai, ML Ralph, T Manly, S Price
- Journal:
- Neuro-Oncology
- Citation info:
- 27(Supplement_5):v328-v328
- Publication date:
- 11th Nov 2025
- Full text
- DOI