The search strategy resulted in a total of 911 records. Of these, 270 records were excluded based on the objective and/or design of the study. Subsequently, another 404 records were excluded based on the recruitment status of the study. As a result, a total of 237 records were classified eligible and included for molecular criteria assessment. See Fig. 1 for an overview of the selection process.

Study selection process. GBM glioblastoma

In 181 (76%) of the 237 included studies, molecular criteria (other than diagnostic) were not included in the study design. Of the remaining 56 studies, at least one specific genomic alteration as an upfront inclusion criterium for study participation was required in 33 (59%) of those studies (Table 1). The remaining 23 (41%) studies applied molecular criteria after patient inclusion, for instance, for drug response correlation. The mean number of study participants in these 33 studies was 38 (range 10–200). The most frequent study phase was 1 (64%, 21/33), followed by phase 2 (24%, 8/33) and phase 1–2 (12%, 4/33). Looking to the in-/exclusion criteria, in most of these studies the glioblastoma recurrence was not specified (73%, 24/33), but was occasionally limited to first (21%, 7/33) or ‘first or second’ (6%, 2/33) recurrence. The requirement that molecular testing was performed on fresh tumor material (i.e. at recurrence) was not provided in most studies. In two studies, fresh material was used (6%, 2/33), while in 8 studies archival (i.e. from primary setting) and/or fresh tissue was used for molecular testing (24%, 8/33). In the remaining studies either archival tissue sufficed (30%, 10/33) or a requirement regarding the moment of molecular testing was not provided (39%, 13/33). Testing was done by next-generation sequencing (NGS) or RNA sequencing (RNAseq) in 8 and 1 of the 33 studies, respectively. IHC, FISH, and sequencing of DNA via tumor in situ fluid (TISF) collection were used in 14, 3, and 1 of the studies, respectively, while the target analysis method was not specified in 11 of the 33 (33%) studies.

Looking somewhat further into detail, EGFR (mutation or amplification, n = 11) was the most frequently investigated gene, followed by CDKN2A/B or C (deletion), CDK4/6 (amplification), and RB (wild-type status), each being investigated in 4 studies. Of the protein targets, B7-H3 and MMP2 were the most frequently (n = 2 each) studied, both in the context of chimeric antigen receptor T-cell (CAR-T) therapy (Table 1). All these alterations were used as a selection criterium for study participation.

Systemic therapies were investigated in 48 of the 56 studies on molecular criteria, but not all these studies required upfront matching based on at least one genomic alteration (Table 2). The majority (n = 27) of these therapeutic studies investigated one or more targeted therapies. Within the targeted therapy group, abemaciclib was the most frequently studied target-matched (CDKN2A/B/C, CDK4/6, RB) drug. Ribociclib, targeting the same genomic alterations, was the second most frequently studied drug. Focusing on immunotherapies, CAR-T therapies were the most frequently studied therapies that, inherently to the principle of CAR-T therapy, required upfront matching based on a genomic alteration. Other therapies being studied in recurrent glioblastoma included acetazolamide and mycophenolate mofetil, both known for potentiating chemosensitivity. In the study on acetazolamide, patients receive concomitant temozolomide, and Bcl-3 expression level is determined to examine the ability of Bcl-3 to predict treatment response. Mycophenolate is studied in combination with temozolomide and/or radiation therapy, and as an exploratory objective, molecular characterization of all glioblastoma tissues by RNAseq is performed.