Objectives To (1) stratify patient subgroups according to their distinct walking gait patterns in end-stage knee osteoarthritis (OA); (2) compare measures of joint loading and proximal tibia subchondral trabecular bone (STB) microarchitecture among these gait subgroups. Design Twenty-five knee OA patients undergoing total knee arthroplasty (TKA) had pre-operative gait analysis. Following surgery, excised tibial plateaus were micro-CT-scanned and STB microarchitecture analysed in four tibial condylar regions of interest. Peak knee moments were input to k-means cluster analysis, to identify subgroups with homogeneous gait patterns. Joint loading and STB microarchitecture parameters were compared among gait subgroups (Kruskal–Wallis, Bonferroni-corrected Mann–Whitney U tests). Results Three gait subgroups were revealed: biphasics (n = 7), flexors (n = 9), counter-rotators (n = 9). Peak knee adduction moment (KAM) and KAM impulse were significantly higher (P < 0.05) in biphasics than in flexors and counter-rotators (KAM = −0.65, −0.40 and −0.21 Nm/kg, respectively), suggesting a higher medial-to-lateral tibiofemoral load ratio in biphasics. Interestingly, STB medial-to-lateral bone volume fraction (BV/TV) ratio was also significantly higher (more than double) in biphasics and flexors than in counter-rotators (2.24, 2.00 and 1.00, respectively), whereas in biphasics it was only 10% higher than in flexors and not significantly so. Conclusions Within the confines of the limited sample size, data suggests that different mechanisms between the biphasic and flexor gait subroups may generate comparable loads upon the tibial plateau and corresponding bony responses, despite significantly lower KAM indices in flexors. Hence, in flexor gait OA patients, conservative treatments designed to reduce KAM, may not be appropriate. Understanding joint loading among walking gait patterns and relationships to bone microarchitecture may aid at identifying/improving management of persons at risk for developing knee OA.