Background: Standard first line chemoirradiation with temozolomide is associated with distinctive peripheral blood immune cell profiles in IDH wildtype glioblastoma. Whether such profiles at recurrence are associated with survival has not been studied in detail. Methods: Peripheral blood mononuclear cells of 21 healthy donors and of 91 patients with IDH wildtype glioblastoma were analyzed by flow cytometry at 1^st recurrence. Patients received either (i) standard chemoirradiation with temozolomide (TMZ) followed by dose-intensified TMZ at first recurrence within the phase II trial DIRECTOR (N = 52) or (ii) hypofractionated radiotherapy with or without bevacizumab (N = 39) followed by investigators’ choice within the phase II trial ARTE. Patients were classified based on unsupervised analyses of PBMC profiles at 1^st recurrence. Associations with survival were explored in multivariate Cox models controlling for established prognostic and predictive factors. Results: At 1^st recurrence, two patient clusters were identified in the DIRECTOR cohort which differed in CD4+ T-cell fractions, but not with respect to CD8+ T-cells, CD4+;CD25+;FoxP3+ regulatory T-cells, B-cells or monocytes. The composition of CD4+, CD8+ or regulatory T-cell fractions was similar in both clusters. All control samples clustered with the CD4_high cluster. Patients in both clusters did not differ by established prognostic factors, including age, O6-methylguanine-DNA-methyl-transferase (MGMT) gene promoter methylation, tumor volume, Karfnosky performance score or steroid use. Progression-free survival was similar (CD4_highvs CD4_low 2.1 vs 2.4 months, p = 0.19), whereas post-recurrence overall survival was longer among the CD4_high cluster (12.7 vs 8.7 months, p = 0.004). At 2^nd recurrence after dose-intensified TMZ re-challenge, monocyte fractions increased, whereas memory CD4+ T-cell fractions decreased. Higher memory CD4+ fractions were associated with longer overall survival at 2^nd recurrence (p = 0.004). The reported associations were retained in multivariate Cox models controlling for established prognostic factors. In the ARTE cohort, CD4+ T cell fractions at 1^st recurrence did not differ compared to diagnosis (p = 0.91) and there were no associations with bevacizumab (p = 0.28) or survival (p = 0.74), supporting that the effects observed in the DIRECTOR cohort were driven by TMZ. Conclusions: We conclude that TMZ-associated memory CD4+ T-cell depletion may have deteriorating effects on the survival of glioblastoma patients.