Patients and baseline characteristics
Four hundred and seventy consecutively enrolled patients with high-risk cancers (expected cure rate lower than 30% assessed by both referring oncologist and central review) were consented for the PRISM study between 14 September 2017 and 31 December 2020. Eighty-six patients were ineligible because of a non-high-risk cancer diagnosis on central review, lack of appropriate sample or death before MTB presentation (Extended Data Fig. 1). Hence, 384 patients discussed at the MTB were included in this analysis. The molecular profile of 181 of these patients has been described previously1. At the time of data cutoff on 30 June 2022, 244 patients were deceased, two were lost to follow-up and the remaining 138 had at least an 18-month follow-up from enrollment (median = 33.7; range: 18.2–56.9 months). The 3-year overall survival (OS) of the 384-patient cohort was 34% (95% confidence interval (CI) = 29–40%) (Fig. 1a).
a, OS of 384 patients with high-risk cancers. b, Number of PGT recommendations per patient. c, Frequency of PGT recommendations according to cancer type. d, Number of PGT recommendations from the highest level of supporting evidence to the lowest (tier 1, clinical evidence in the same cancer; tier 2, clinical evidence in a different cancer; tier 3, preclinical evidence in the same cancer; tier 4, preclinical evidence in a different cancer; tier 5, consensus opinion). e, Distribution of PGT recommendation tier according to cancer type. f, Types of targeted therapy in relation to the drug target. Targeted therapy was categorized into targeted monotherapy, targeted dual therapy and targeted agent in combination with chemotherapy. The corresponding molecular pathway for each of the drug targets is shown. c, P values for comparison of proportions using a two-sided chi-squared test.
Of the 384 eligible patients, 160 patients were enrolled at first cancer diagnosis, 184 patients at first relapse and 40 patients after two or more previous relapses (Extended Data Table 1 and Supplementary Data 1). The cohort consisted of 146 central nervous system (CNS) tumors, 183 solid tumors and 56 hematologic malignancies (HMs). One patient with a germline mutation in the TP53 gene had two synchronous tumors analyzed (medulloblastoma (MB) and osteosarcoma (OST)). Median age at enrollment was 10.9 years (range 0.1–46 years), including 14 adults (aged older than 21 years) with pediatric-type cancers.
All patients had at least one somatic NGS assay performed. Both WGS and whole-transcriptome sequencing (WTS) were successfully conducted on 319 of 385 samples (83%). WGS alone was performed on 54 samples, targeted panel on ten and targeted panel plus WTS on two cases either because of insufficient DNA or RNA or because only formalin-fixed paraffin-embedded tissue was available (Extended Data Table 1). DNA methylation profiling was performed in 298 of 329 CNS tumors or sarcomas. Germline WGS was performed on 374 patients and germline targeted panel on ten patients.
Identification of therapeutic targets
Molecular findings were classified as reportable or actionable as described previously1 and discussed in the national MTB. A five-tier system was used to assign the strength of the PGT recommendation (Methods and Supplementary Data 2). PGT was recommended only if age-specific drug safety data were available and there was a possibility of drug access in Australia via registered indication, clinical trials, compassionate access or off-label use. Two hundred and fifty-six patients (67%) received at least one PGT recommendation, with a total of 510 PGT recommendations made (Fig. 1b). The recommendation rate was significantly higher for CNS tumors than solid tumors (73% versus 62%; P = 0.048) (Fig. 1c). While 53% of the recommendations had supporting clinical evidence (tiers 1 and 2), 43% were derived from preclinical evidence (tiers 3 and 4) (Fig. 1d). CNS tumors had significantly fewer tier 1 recommendations compared with solid tumors (14% versus 25%; P = 0.007) and HMs (14% versus 36%; P < 0.0001) (Fig. 1e). The 510 PGT recommendations consisted of 74% targeted monotherapy, 12% targeted dual therapy, 13% targeted and chemotherapy combination, and 1% chemotherapy alone. Therapies targeting the phosphoinositide 3-kinase (PI3K)/mammalian target of rapamycin (mTOR) (20%) and mitogen-activated protein kinase (MAPK) (15%) pathways were most frequently recommended, followed by poly(ADP-ribose) polymerase (PARP) (10%) and cyclin-dependent kinase 4 (CDK4) and CDK6 inhibitors (8%) (Fig. 1f). Of the receptor tyrosine kinases (RTKs), fibroblast growth factor receptor (FGFR) (28%) was the most common target followed by vascular endothelial growth factor (VEGF)/vascular endothelial growth factor receptor (VEGFR) (20%) and epidermal growth factor receptor (EGFR)/ERBB (16%).
Clinical uptake and drug access for PGT
Of the 256 patients with a PGT recommendation, 110 (43%) were subsequently treated with a PGT, with a median time of returning results at 6.6 weeks. Seventy percent of PGTs were commenced within 3 months of the MTB, with a median start time of 9 weeks (1 day–2.5 years). Three patients started treatment before MTB after rapid communication of results to the treating clinician. In total, 117 PGTs were administered to 110 patients, with six patients receiving two or more consecutive PGTs. The early clinical responses to 37 of these PGTs were reported previously1. Of note, clinical testing for the specific PGT target was only available for 13 targets and was performed in ten. For these ten patients, only six returned a positive test, with four returning a negative or equivocal result. Thus, for 95% of patients receiving PGT, their driver either could not be or was not detected through locally available testing (Supplementary Table 1).
The mechanism according to which patients gained access to the 117 PGTs included compassionate access in 42 (36%), funding via the local clinical institution in 39 (33%), clinical trial enrollment in 19 (16%), funding from the government Pharmaceutical Benefit Scheme in ten (9%), cost sharing arrangement between hospital and drug company in five (4%) and self-funded in two (2%) (Supplementary Data 3).
Clinical benefit of PGT
Of the 117 administered PGTs, 99 (received by 93 patients) were eligible for outcome analysis (Supplementary Data 3). Eighteen PGTs were excluded from the analysis, including 14 patients whose treatment duration was less than 4 weeks (Extended Data Fig. 1). Disease responses were evaluated using the Response Evaluation Criteria in Solid Tumors (RECIST), Response Assessment in Neuro-Oncology (RANO) or Positron Emission Tomography Response Criteria in Solid Tumors (PERCIST) criteria. Measurable disease was present at the start of 70 PGTs, with complete responses (CRs) observed in six (9%), partial responses (PRs) in 19 (27%), stable disease (SD) in 24 (34%) and progressive disease (PD) in 21 (30%) (Fig. 2a–c). The OR rate (ORR) (CR or PR) was similar for CNS and solid tumors (35% versus 34%). In addition, 20 PGTs (19 patients) were commenced for evaluable but non-measurable disease, with two CRs, ten SDs and eight PDs. Thus, the outcome for 90 evaluable PGTs (70 measurable and 20 non-measurable, excluding nine with no evidence of disease at the start of PGT) was CR in 9%, PR in 21%, SD in 38% and PD in 32% (Fig. 2d).
a, Waterfall plot for 31 CNS tumors with measurable disease at the start of a PGT. Treatment response was evaluated using the RANO criteria. The dotted lines at 25% and −50% delineate the category of response (≥25%, PD; 25 to −50%, SD; −50% or lower to −99%, PR; −100%, CR). b, Waterfall plot for 35 solid tumors with measurable disease at the start of a PGT. Treatment response was evaluated using the RECIST or PERCIST criteria. The dotted lines at 25% and −30% delineate the category of response (≥25%, PD; 25 to −30%, SD; −30% or lower to −99%, PR; −100%, CR). c, Response according to cancer type to 70 PGTs given with measurable disease. d, Response according to cancer type to 90 PGTs with evaluable disease (70 measurable and 20 non-measurable). e, OCB rate in 97 PGTs. OCB was defined as CR, PR and SD of 24 weeks’ duration or longer. f, PFS ratio for 31 PGTs. The PFS ratio was defined as the PFS duration of the PGT to that of a previous treatment in the same patient. A PFS ratio greater than 1.3 (above the dotted line) represents prolongation of the progression-free period for more than 30% by the PGT compared to a previous treatment. The color of the dots denotes the clinical course. g,h, PFS (g) and OS (h) stratified according to PFS ratio. A two-sided log-rank test was used to compare the Kaplan–Meier survival curves.
The duration of disease control can be meaningful for patients with high-risk cancers receiving new therapies; objective clinical benefit (OCB) (CR, PR and sustained SD for 24 weeks or longer) has been used as an endpoint in clinical trials of targeted agents10,11. Therefore, we evaluated OCB for 97 PGTs, including nine PGTs commenced with no evidence of disease (Extended Data Fig. 1). OCB was observed in 55% (53 of 97) of PGTs and was similar across tumor types (Fig. 2e).
The intra-patient progression-free survival (PFS) ratio has been used to compare the efficacy of PGT with previous treatments for the same patient, with clinical benefit defined as a PFS ratio greater than 1.3 (refs. 12,13). Thirty-one patients treated with PGT were assessable for PFS ratio and 42% (95% CI = 25–61%) had a PFS ratio greater than 1.3 (Fig. 2f). To determine whether a prolonged PFS ratio correlated with improved survival, we compared patients with a PFS ratio greater than 1.3 with those with a ratio of 1.3 or lower and found that they had a significantly improved PFS (2-year PFS 36% versus 0%; P = 0.02) (Fig. 2g). There was a similar difference in OS that did not reach statistical significance (2-year OS 46% versus 8.3%; P = 0.10) (Fig. 2h).
PGT improved outcomes compared to other treatments
To understand whether PGT improved outcomes compared with other therapies, we next compared the outcomes for patients who received PGT versus non-PGT, that is, other therapies not recommended by the MTB, including standard of care (SOC) treatment and new or targeted therapies not guided by molecular findings and not recommended by the MTB, termed unguided therapy (UGT) in this study. One hundred and seventy-three patients whose treatment commenced after MTB and were evaluable for disease progression (treatment duration 4 weeks or longer and progression-free for 4 weeks or longer) were included in the survival analysis. Eighty-nine and 84 patients received a PGT or non-PGT as first treatment after MTB, respectively, and were compared for OS. For the PFS analysis, 99 PGTs were compared with 132 non-PGTs (75 SOC, 45 UGT and 12 other experimental treatments). Treatment with PGT resulted in significantly improved PFS when compared with non-PGT (2-year PFS 27% versus 11%; P = 0.01) (Fig. 3a), whereas the difference in OS did not achieve statistical significance (2-year OS 38% versus 24%; P = 0.08) (Fig. 3b), perhaps because of different salvage therapies (Supplementary Table 2).
a,b, PFS (a) and OS (b) stratified according to PGT and non-PGT commenced at any point after MTB discussion. c,d, PFS (c) and OS (d) stratified according to PGT and UGT, that is, new therapy not molecularly guided. e, Response to PGT and UGT in patients with evaluable disease. f, OCB rate in PGT and UGT. OCB was defined as CR, PR and SD of 24 weeks’ duration or longer, and ongoing CR of 24 weeks or longer for patients who were in CR at the start of treatment. A two-sided chi-squared test was used to compare the CR and PR rate in e and OCB rate in f. g,h, PFS (g) and OS (h) stratified according to PGT and SOC. For OS comparison, a patient was categorized according to the first treatment that was initiated after MTB discussion. The Kaplan–Meier survival curves were compared using a two-sided log-rank test.
We asked whether disease status could impact treatment. Of 99 PGTs, 49 were given before disease progression since study enrollment, 42 after one disease progression and eight after two or more episodes. Of 132 non-PGTs, 22 were given before disease progression since study enrollment, 76 after one disease progression and 34 after two or more episodes. For treatments given after no or one episode of progression, the 2-year PFS was 28% for PGT and 14% for non-PGT (P = 0.07). For treatments received after two or more disease progressions, there was no difference in PFS (2-year PFS 0% versus 3%; P = 0.47) (Extended Data Fig. 2).
It is possible that PGT was superior to the other therapies as these patients received new agents rather than standard cytotoxic therapies. Therefore, we compared the outcomes for PGT with UGT. Instances of patients receiving UGT included those enrolled on phase I trials of agents not requiring biomarkers, or treatments based on previous clinical trial data, for example, pazopanib for sarcoma and venetoclax for leukemia. A total of 45 UGTs were commenced after MTB in 36 patients (Extended Data Table 2 and Supplementary Data 4). Two UGTs were excluded from the response evaluation because the patients were disease-free at the start of treatment. PGT resulted in significantly improved PFS compared with UGT (2-year PFS 26% versus 5.2%; P = 0.003) (Fig. 3c), whereas the difference in OS did not achieve statistical significance (2-year OS 38% versus 20%; P = 0.15) (Fig. 3d). The response rate (CR/PR) was significantly higher after PGT compared with UGT (30% versus 2.3%; P < 0.0001) (Fig. 3e). Similarly, a higher OCB rate was observed with PGT versus UGT (55% versus 23%; P = 0.0003) (Fig. 3f).
SOC treatment was commenced after MTB discussion in 65 patients. SOC was defined as treatment routinely used in a tumor type or treatment reported to have proven clinical activity, for example, irinotecan (IRN) or temozolomide (TMZ) for relapsed Ewing’s sarcoma (EWS) and FLAG-Ida (fludarabine, cytarabine, idarubicin, granulocyte-colony stimulating factor) for relapsed acute leukemia. Fifty-five of these patients received SOC as first treatment after MTB and were included in the SOC group for OS comparison. A total of 75 SOC regimens were evaluated for PFS. We found that PGT led to significantly improved PFS compared to SOC (2-year PFS 26% versus 12%; P = 0.049) (Fig. 3g), whereas the difference in OS did not achieve significance (2-year OS 38% versus 23%; P = 0.11) (Fig. 3h).
Three patients with pilocytic astrocytomas with atypical aggressive clinical courses and multiple previous disease progressions were treated with PGT. Targetable molecular findings included two with protein tyrosine phosphatase non-receptor type 11 (PTPN11) and FGFR1 mutation and a phosphotyrosine interaction domain-containing protein 1-BRAF fusion (Supplementary Data 5). Because pilocytic astrocytoma could have prolonged PFS compared to other tumor types in the cohort, we repeated the analysis excluding these three cases. We found that the improvement in PFS, OR and OCB remained statistically significant when comparing PGT with all other treatments (Extended Data Fig. 3).
Factors predicting response to PGT
To better understand the characteristics of patients who received a PGT, we examined the target genes, category of molecular aberrations, strength of evidence, tumor type and type of response for each individual patient. We observed responses or prolonged SD across all different scenarios (Fig. 4a,b). Details of CNS tumors with OCB to PGT are provided in Supplementary Table 3. Patients for whom PGT did not lead to OCB are shown in Extended Data Fig. 4. We found that every tier was associated with clinical responses, with response rates highest for tier 1 (39%), but not significantly different to tier 2 (18%; P = 0.07) or tiers 3–5 (31%; P = 0.49) (Table 1 and Extended Data Fig. 5a). Only one of three patients benefited from tier 5 PGT: a patient with ependymoma (EPN) with high vascular endothelial growth factor A (VEGFA) RNA expression with PR to bevacizumab. The remaining two patients progressed rapidly—an H3K27M mutant diffuse midline glioma with PDGFRA mutation treated with ponatinib and then regorafenib, and an MB with FGFR3 mutation given pazopanib, ifosfamide and doxorubicin. Similarly, OCB was also highest for tier 1 (74%) compared with either tier 2 (41%; P = 0.008) or tier 3–5 (44%; P = 0.01) (Extended Data Fig. 5a). These results translated to improved survival, with tier 1 PGT resulting in longer PFS (tier 1 versus 2; P = 0.07 and tier 1 versus tiers 3–5; P = 0.001), and OS (tier 1 versus 2; P = 0.03 and tier 1 versus tiers 3–5; P = 0.0003) when compared to other tiers (Fig. 5a and Table 1), and longer PFS compared to non-PGT (P = 0.0002).
a,b, Swimmer plot of 53 PGTs (51 patients) leading to OCB. OCB includes CR, PR and SD for a duration of 24 weeks or longer and ongoing CR of 24 weeks or longer for patients who were in CR at the start of treatment. Forty-one PGTs with tier 1 and 2 recommendations are show in a and 12 with tier 3–5 recommendations are shown in b. The color of the bars indicates the tier of a PGT recommendation. The symbols indicate the responses and treatment status. The diagnosis and molecular targets for each patient are shown. The types of molecular aberration are denoted by different colored text. The fusion or structural variant (SV) is shown in blue, SNVs in red, high RNA expression in green, copy number variant in brown and other alterations in black. AS, angiosarcoma; B-ALL, B cell acute lymphoblastic leukemia; CCM, clear cell meningioma; CET, CNS embryonal tumor not otherwise specified; DMG, diffuse midline glioma H3K27M-altered; DSRCT, desmoplastic small round cell tumor; ERMS, embryonal rhabdomyosarcoma; GIST, gastrointestinal stromal tumor; GO, glioma other; HCC, hepatocellular carcinoma; HGG, high-grade glioma; IFS, infantile fibrosarcoma; MPNST, malignant peripheral nerve sheath tumor; MTV, medullary thyroid carcinoma; NB, neuroblastoma; SPNP, solid pseudopapillary neoplasm of the pancreas; T-ALL, T cell acute lymphoblastic leukemia; US, undifferentiated sarcoma; WT, Wilms tumor.
a–d, PFS and OS stratified according to the tier of PGT (a), the types of molecular aberration (b), PD from enrollment to the start of PGT (c) and the number of favorable prognostic factors (d). A two-sided log-rank test was used to compare the Kaplan–Meier survival curves of two groups; the reference subgroup is indicated by a dash.
We next assessed response and survival based on the PGT target. Patients whose treatment targeted a fusion or SV had the highest response rate (CR/PR) of 60%, compared with single-nucleotide variants (SNVs) (32%; P = 0.07), high RNA target expression only (15%; P = 0.006) and copy number variation (CNV) (14%; P = 0.01) (Extended Data Fig. 5b and Table 1). A similar trend was observed for the OCB. These results correlated with survival outcomes. The 2-year PFS was superior for PGT targeting a fusion/SV (68%), compared with SNV (30%; P = 0.057), high RNA expression alone (5.9%; P = 0.002) or CNV (7.7%; P = 0.003) (Fig. 5b and Table 1). Evaluation of OS demonstrated a similar trend with PGT targeting a fusion/SV leading to a 2-year OS of 69%. PGT targeting a fusion/SV and SNV also led to improved PFS when compared with non-PGT (P = 0.001 and P = 0.02, respectively).
Of note, 46% (11 of 24) of PGTs targeting high RNA expression alone (not associated with SV, SNV or CNV) led to OCB (Extended Data Table 3), including three ORs. This included a solid pseudopapillary neoplasm of pancreas (everolimus for high Ras homolog enriched in brain (RHEB)), EPN (bevacizumab for high VEGFA) and rhabdomyosarcoma (temsirolimus/vinorelbine/cyclophosphamide for high AKT2). Genes within the AKT/mTOR, VEGF/VEGFR and FGF/FGFR pathways were most frequently targeted for high RNA expression (Extended Data Table 3 and Fig. 5c), with a similar OCB rate (50–60%).
Of the 99 PGTs administered, 57 were administered as targeted monotherapy, 18 as dual targeted therapy, 21 as combination targeted and chemotherapy and three as chemotherapy. There was no difference in 2-year PFS between targeted monotherapy and dual therapy (32% versus 31%; P = 0.6) (Extended Data Fig. 6a). Similar response rates (31% versus 29%; P = 0.75) and OCB rates (52% versus 55%; P = 0.84) were observed for targeted agents administered in combination with chemotherapy and targeted monotherapy and dual therapy (Extended Data Fig. 6b); however, targeted chemotherapy was associated with significantly inferior survival when compared with targeted monotherapy and dual therapy (2-year PFS 0% versus 32%; P = 0.03 and 2-year OS 15% versus 42%; P = 0.048) (Table 1 and Extended Data Fig. 6c).
To evaluate the optimal time to initiate PGT, we assessed patients’ disease status at the start of treatment and found a significant correlation with clinical outcome. From enrollment, patients receiving PGT before relapse or progression had a significantly higher response (40% versus 20%; P = 0.04) and OCB rates (74% versus 36%; P = 0.0001) (Extended Data Fig. 5d). This translated to better survival compared to patients receiving PGT after subsequent disease progression (2-year PFS 42% versus 12%; P < 0.0001 and 2-year OS 53% versus 29%; P = 0.0002) (Fig. 5c and Table 1) and patients receiving non-PGT (2-year PFS 12%; P < 0.0001).
There was no significant difference in outcome between tumor types treated with PGT except for a significantly poorer outcome for patients with HM (P = 0.02) (Extended Data Fig. 6d,e). As expected, the uptake of PGT in HM was low (22%), and PGT was given to heavily pretreated patients because of the availability of effective previous salvage therapies. Of the eight patients with HM who received a PGT, three were not evaluable as they were treated for fewer than 4 weeks. Three evaluable patients progressed rapidly and two patients (NUP214-ABL1 fusion and NR3C1 monoallelic loss) responded (to dasatinib and venetoclax/navitoclax, respectively) and proceeded to transplant.
Finally, we conducted a multivariate analysis of prognostic factors. Tier 1 evidence, fusion/SV, PGT given before relapse or disease progression, and non-hematological malignancy had independent prognostic significance for PFS, with hazard ratios of 0.43, 0.42, 0.50 and 0.21, respectively (Extended Data Table 4). We next asked whether combinations of independent favorable factors impacted the outcomes for PGT. We identified whether each patient had 0, 1, 2 or 3 of each of the following: tier 1 evidence; fusion/SV; and PGT given before PD. HMs were excluded because of very small numbers. We found that the number of favorable factors was significantly associated with improved response rate, OCB, PFS and OS (Fig. 5d, Extended Data Fig. 5e and Table 1). Patients with three favorable factors demonstrated the highest response rates and OCB of 75% and 100%, respectively. The 2-year PFS was 6.7% for patients with no favorable factors, 21% for patients with one factor (P = 0.03), 43% for those with two factors (P < 0.0001) and 88% for those with three favorable factors (P = 0.0004). The same trend was observed for OS.