Ganetespib in Combination with Pemetrexed–Platinum Chemotherapy in Patients with Pleural Mesothelioma (MESO-02): A Phase Ib Trial

This article is featured in Highlights of This Issue, p. 4713

Malignant pleural mesothelioma (MPM) is an incurable, rapidly lethal cancer arising most commonly from the parietal pleural mesothelium, and is associated with exposure to asbestos. Although the number of deaths due to MPM has increased globally (1) there has been no new licensed therapy since 2004. The combination of an antifolate and platinum agent is an effective first-line treatment for MPM. Pemetrexed and cisplatin, the approved standard, has a response rate of 41.3%, with a median progression free survival (PFS) of 5.7 months, and median overall survival (OS) of 12.1 months (2). Using carboplatin instead of cisplatin has comparable activity (3), and a platinum combination with another antifolate raltitrexed is also effective (4).

There remains an unmet clinical need for new, effective therapies that can improve outcomes in the first-line treatment setting. Addition of the VEGF inhibitor bevacizumab to pemetrexed-cisplatin therapy can improve overall survival (18.8 months with bevacizumab vs. 16.1 months without; HR = 0.77, P = 0.0167) and progression-free survival (9.2 months with bevacizumab vs. 7.3 months without; HR = 0.61, P < 0.0001), but is not licensed (5), and is only recommended by National Comprehensive Cancer Network guidelines to be used in unresectable patients who are able to receive bevacizumab (6). No positive randomized controlled studies have shown improved survival in either the maintenance (7) nor relapsed treatment settings (8–10). A recent open label phase II trial did however show improvement in progression-free survival with switch maintenance gemcitabine (11).

Hsp90 is a molecular chaperone that mediates posttranslational stabilization of critical oncogenic signaling molecules, via a repertoire of client proteins that include oncogenic kinases relevant to MPM such as AXL and MET (12). Hsp90 inhibition has been reported to induce apoptosis in MPM via an MCL1-dependent mechanism (13) and facilitates the evolution of drug resistance (14). Acquisition of aneuploidy has been reported as a mechanism of resistance to Hsp90 inhibition (15).

Thymidylate synthase (TS) is a Hsp90 client, implicated in antifolate resistance, which is downregulated following inhibition of Hsp90 (16). Furthermore, preclinical studies show that inhibition of Hsp90 mediates synergistic toxicity due to cisplatin (17). Ganetespib (ADX-1612), an Hsp90 inhibitor, is a synthetic quadricyclic triazolone with a small molecular weight that binds to the adenosine triphosphate pocket in the N-terminus of Hsp90 (17, 18). Single-agent ganetespib demonstrates significant activity for downregulating Hsp90 client protein levels with acceptable toxicity at a recommended dose of 200 mg/m² from phase II studies (18, 19). Furthermore, inhibition of Hsp90 with ganetespib has been shown to enhance T-cell–mediated antitumor immune response (20). We hypothesized that the addition of ganetespib to pemetrexed and either cisplatin or carboplatin, can be safely delivered, that there might be a synergistic interaction clinically, and that patients harboring genomic instability (reflected in somatic copy number alterations, loss of heterozygosity, and homozygous deletions) might exhibit resistance to ganetespib.

MESO-02 (ClinicalTrials.gov identifier: NCT01590160) was a multicenter phase I/II study of first-line ganetespib with pemetrexed/platinum, in patients with MPM. Here, we report the results of the phase Ib stage, in which the primary objective was to find a safe dose of ganetespib when combined with standard platinum and pemetrexed. A major secondary objective was to examine clinical efficacy. The trial was conducted in accordance with the World Medical Association Declaration of Helsinki and approved by the UK Medicines and Healthcare Products Regulatory Agency (clinical trial authorization number: 20363/0317/001-0001), the Research Ethics Service Committee East Midlands, Derby (REC reference no. 12/EM/0448), and the research and development department of each participating National Health Service trust. All patients provided written informed consent.

Key inclusion criteria included: age 18 years or older; histopathologic confirmation of MPM, with measurable disease using meso-modified Response Evaluation Criteria in Solid Tumors (RECIST) v1.0 (21); an Eastern Cooperative Oncology Group (ECOG) performance status of 0 to 1; adequate hematologic status and organ function; and chemotherapy naïve. Key exclusion criteria included: evidence of CNS metastases that require local treatment prior to systemic cytotoxic chemotherapy; receipt of extensive radiotherapy (except drain-site radiotherapy), systemic chemotherapy, or other antineoplastic therapy within 4 weeks before enrollment; uncontrolled intercurrent illnesses; known serious cardiac illness; and history of prior gastrointestinal illness. Detailed eligibility criteria are given in the study protocol (see Supplementary Appendix).

Patients were given a 1-hour intravenous ganetespib infusion on days 1 and 15 of each 21-day cycle, at one of three dose levels: 100 mg/m², 150 mg/m², or 200 mg/m². Patients also received a 10-minute intravenous pemetrexed infusion of 500 mg/m² (with vitamin B₁₂ and folate supplementation) immediately after ganetespib infusion on day 1 only. All patients received either cisplatin (75 mg/m² i.v. over 2 hours) or carboplatin (AUC5 intravenously over 30 minutes), 30 minutes after the completion of pemetrexed infusion. Patients in the trial were initially only allowed cisplatin, and once the safety profile was shown to be acceptable, carboplatin was allowed subsequently (at the clinician's discretion and influenced by expected patient tolerability). Having either platinum agent was incorporated in the trial to reflect routine practice.

Within the MESO-02 study there were separate cohorts for cisplatin or carboplatin, to ensure acceptable safety of the combination of ganetespib and platinum therapy specifically in patients with MPM.

For patients receiving the ganetespib–pemetrexed–cisplatin triplet (i.e., the cisplatin cohort), dose escalation of ganetespib was conducted using the 3 + 3 design with a starting dose of 100 mg/m². In each cohort, if no dose-limiting toxicities (DLT) were observed, recruitment proceeded to the next cohort of 3 patients. If there was one DLT, the cohort involved was expanded to 6. If there were no further DLTs in the cohort of 6, the next cohort was administered ganetespib with chemotherapy at the next highest dose. If 2 or more DLTs were observed, in 3 or 6 patients at a given dose, dose escalation would discontinue and no higher dose considered. The cohort at the estimate of the MTD was then expanded to 9 patients overall. The maximum planned sample size for the cisplatin cohort was 27 patients.

For patients receiving carboplatin with ganetespib and pemetrexed (i.e., the carboplatin cohort), dose escalation of ganetespib was conducted using an accelerated titration design with a starting dose of 100 mg/m². At dose levels below 200 mg/m², one patient would receive treatment. If no DLT was observed, the next patient would receive the next highest dose; otherwise, a 3 + 3 design would begin (i.e., the same as for the cisplatin-treated cohort). If ganetespib reached the estimate of the MTD, the cohort was expanded to 9 patients overall. An accelerated titration design was used here as the carboplatin cohort was introduced following a protocol amendment after 9 patients had been treated with cisplatin, none of whom experienced any DLTs. An accelerated escalation design toward 200 mg/m² ganetespib struck a balance between quickly moving toward a likely safe dose, while still allowing for reintroduction of a 3 + 3 procedure if any DLTs were observed. The maximum planned sample size for the carboplatin cohort was 18 patients.

Patients who completed 6 cycles of chemotherapy without signs of disease progression in either cohort could go on to receive ganetespib as maintenance monotherapy, using the same dose they had already been given. Ganetespib would be given on days 1 and 15 of each 21-day cycle, and continued until toxicity, progression, or patients decided to stop.

Collection of archival formalin-fixed paraffin-embedded (FFPE) diagnostic tissue was mandatory and undertaken at each patient's screening visit prior to registration. Hematologic profiling comprised assessment of the hematocrit, hemoglobin, red cell count (RCC), white cell count with differential, and platelets. Biochemical profiling comprised sodium, potassium, urea, creatinine, chloride, bicarbonate, bilirubin, alkaline phosphatase (ALP), aspartate aminotransferase (AST) or alanine aminotransferase (ALT), lactate dehydrogenase (LDH), albumin, total protein, calcium, phosphate, magnesium, and urinalysis (pH, protein, blood, ketones, glucose). These were conducted before each treatment cycle (full blood count and biochemistry, also before day 15 of cycles 1 and 2; urinalysis, also before day 15 of cycle 1). Patients underwent a CT scan of the chest and abdomen for disease response assessment within 28 days of registration, after chemotherapy cycles 2, 4, and 6, then every 6 weeks for 12 months. Scans were assessed according to meso-modified RECIST v1.0 (21).

Toxicities were graded using the NCI's Common Terminology Criteria for Adverse Events (NCI CTCAE v4.0). A DLT was defined as any of the following adverse events deemed definitely, probably, or possibly related to ganetespib therapy: grade 3 or 4 nonhematologic events except diarrhea, nausea, and vomiting lasting more than 48 hours despite maximum medical therapy; grade 4 thrombocytopenia or neutropenia lasting longer than 7 days; febrile neutropenia; any drug-related adverse event leading to an interruption of ganetespib for longer than 14 days; or any clinically significant toxicity leading to dose reduction for ganetespib. DLT assessment applied to cycles 1 and 2 only for patients in the cisplatin cohort, and cycle 1 only for the carboplatin cohort. Dose escalation decisions and DLT reviews were made by members of the Trial Management Group.

Efficacy outcomes were progression-free survival (PFS), the time from study registration to confirmed disease progression or death by any cause and overall survival (OS), death by any cause, and best response rate [defined as the percentage of patients with a best overall response of complete response (CR) or partial response (PR) as per meso-modified RECIST v1.0].

For DNA copy number analyses, genomic variables of interest were somatic copy number alterations (SCNA; number of somatic changes to chromosome structure that lead to gain or loss in copies of sections of DNA), global loss of heterozygosity (LOH; the number of somatic cells containing only one copy of an allele), and total homozygous deletions (total number of biallelic copy number losses). These were assessed in the baseline (diagnostic) tumor sample; 80 ng of DNA were extracted from archival diagnostic FFPE tissue blocks with the QIAmp DNA Mini Kit (Qiagen) and analyzed using the OncoScan FFPE Assay Kit (Thermo Fisher Scientific), which utilizes molecular inversion probe (MIP) technology. MIP probes were added to the FFPE DNA for annealing performed for 16–18 hours. Gap fill reaction was then performed. Uncircularized MIP probes and genomic DNA were digested and circular MIP probes were linearized and amplified by PCR. Following a second round of PCR amplification, amplicons were cleaved into smaller DNA fragments with the HaeIII enzyme to improve sample hybridization onto the OncoScan arrays. Samples were allowed to hybridize for 16–18 hours. After hybridization, arrays were stained and washed and loaded into the GeneChip Scanner (22). BioDiscovery Nexus Express for OncoScan software was then used to define copy number alterations and loss of heterozygosity. The software uses the TuScan algorithm to generate an estimate of tumor ploidy and aberrant cell fraction at each copy number change. Samples were analyzed retrospectively following completion of study enrollment, although blinded to knowledge of both safety and efficacy outcomes from the patients themselves.

The MTD of ganetespib plus pemetrexed and platinum therapy was determined by the dose-escalation design and expansion phase in each platinum therapy cohort. Kaplan–Meier methods were used to analyze PFS and OS. For exploratory genomic analyses, time to progression and associations with genomic variables (SCNA, LOH, and total homozygous deletions at baseline) were assessed using separate Cox regressions, with each genomic variable included as a covariate and time to progression as an outcome (patients with no confirmed progression date were censored at last date known to be alive). Spearman correlation was used to assess the association between best percentage change in total tumor burden (TTB, defined as the sum of six pleural measurements in millimeters determined by CT scan as per modified RECIST) from baseline and genomic variables. All analyses were conducted on data frozen on February 4, 2019 using Stata version 15.1 (23).

Between September 4, 2013 and November 10, 2015, 27 chemotherapy-naïve patients with a confirmed diagnosis of MPM (25 male, 2 female) were recruited. Patient characteristics are shown in Table 1. Median age was 66 years (range 37–76). Nonepithelioid MPM accounted for 22% of patients. Twenty-one patients (78%) had an ECOG PS 1. Only two patients were deemed to have a good prognostic score based on the European Organisation for Research and Treatment of Cancer (EORTC) prognostic scoring system (24). There were 16 patients in the cisplatin cohort and 11 in the carboplatin cohort.

Patients in the cisplatin cohort were administered a median of 4 (range 1–6) treatment cycles (ganetespib and chemotherapy). Patients in the carboplatin cohort were administered a median of 3 (range 1–5) treatment cycles (see Supplementary Table S1). Five patients in the carboplatin cohort (45%) chose to withdraw from the study, with two doing so after experiencing adverse events; full reasons for treatment discontinuation/withdrawal from the study are provided in Supplementary Table S2. Maintenance ganetespib therapy was received by a total of 14 patients with a median of 2 cycles (range 1–60), and 2 patients had ≥10 cycles.

Supplementary Table S3 shows the DLTs observed in each ganetespib dose cohort. At the 200 mg/m² ganetespib dose level, when the cohort was expanded from 3 to 9 patients, one patient in the cisplatin cohort had a DLT (grade 3 toxicity comprising nausea lasting >48 hours). In the carboplatin cohort at 200 mg/m² ganetespib, 1 patient in the first three treated experienced DLT (grade 2 infusion-related reaction). An additional 3 patients were recruited to this dose level, and no more DLTs were observed. The cohort was expanded to 9 patients, with one patient experiencing DLT (grade 3 nausea). Given the observed DLTs, 200 mg/m² was considered to be the MTD.

Grade 3 and 4 toxicities are summarized in Table 2. There were no grade 4 toxicities at the 100 mg/m² ganetespib dose level in either the cisplatin or carboplatin cohorts. At the 150 mg/m² dose in the cisplatin cohort, one patient experienced grade 4 hearing impairment. The most common grade 3 and 4 adverse effects were all related to anemia (6 patients, 22%), decreased neutrophil counts (4 patients, 15%), and nausea/vomiting (4 patients, 15%).

Four patients in the cisplatin cohort had reduced cisplatin (median 56 mg/m²; range 37–56) for a median of 1 cycle (range 1–2). No dose reductions in ganetespib were required in the cisplatin cohort. In the carboplatin cohort, 4 patients had reduced carboplatin (median 405 mg/m²; range 250–480), all for 1 cycle. One patient had their ganetespib dose reduced from 150 mg/m² to 112 mg/m² (1 of 2 treatment cycles), and another patient assigned to receive 200 mg/m² ganetespib did not receive ganetespib on day 15 of cycles 1 and 2 due to hematologic toxicities. Dose reductions for pemetrexed occurred in 4 patients in the carboplatin cohort (reduction from 500 mg/m² to 375 mg/m²) and occurred for a median of 1 cycle (range 1–2).

Median observed follow-up time for all patients was 10.7 months (range 2.3–49.4). Median follow-up time was 12.3 months (range 3.6–49.4) in the cisplatin cohort, and was 8 months (2.3–20.8) in the carboplatin cohort. Overall, 23 patients were evaluable for response (i.e., had at least one evaluable response assessment conducted following treatment). Partial response (≥30% reduction in total tumor burden from baseline) was observed in 14 of 27 patients [objective response rate (ORR) 52%; 95% CI, 32%–71%], with 10 of 18 patients (56%; 95% CI, 31%–78%) receiving 200 mg/m² of ganetespib achieving a partial response (Fig. 1). When only based on patients with evaluable disease, the ORR was 61% (14/23). Twenty-two of 27 patients (81%) had disease control (partial response or stable disease), and one patient had progressive disease. For the 15 evaluable patients treated at the MTD, all (100%) had disease control. Patients with nonepithelioid histology had a lower ORR (1/6; 17%) compared with patients with epithelioid histology (13/21; ORR 62%; 95% CI, 38%–82%). Five of 6 patients with nonepithelioid histology received lower ganetespib doses (100 mg/m² or 150 mg/m²). However, all but 1 of the 6 patients with nonepithelioid MPM had disease control, stable disease, or better (<20% increase in tumor burden). One patient receiving the MTD of ganetespib exhibited a prolonged response, reaching its nadir after 12 months, and was sustained for over 40 months.

Nineteen patients had disease progression, and 21 of 27 patients died. Sixteen patients progressed and died later, and 5 patients died without prior reported progression, the causes of which were MPM. All 5 of these patients had withdrawn from the trial due to patient choice of treatment delay >28 days and were followed up only for assessing overall survival. In all patients, the median PFS was 5.8 months (95% CI, 5.0–8.0; Supplementary Fig. S1), although was not significantly different between the different platinum therapies [median PFS of 5.8 months in both cohorts; HR for cisplatin vs. carboplatin 0.81 (95% CI, 0.34–1.91, P = 0.63); Supplementary Fig. S2]. As with tumor volume, PFS was shown to improve with increasing doses of ganetespib, although not statistically significant (log-rank test for trend, P = 0.10; Supplementary Fig. S3). In patients who had 200 mg/m², the median PFS was 6.3 months (95% CI, 5.0–10.0). Median overall survival was 11.5 months (95% CI, 8.0–19.5; Supplementary Fig. S4), but differed across platinum therapies (cisplatin = 14.4 months; 95% CI, 6.3–28.7; carboplatin = 10.6 months; 95% CI, 6.3–19.5; Supplementary Fig. S5). Overall survival at the MTD was 16.3 months (95% CI, 8.0–21.7; Supplementary Fig. S6).

Aneuploidy has been shown to be associated with acquired resistance to Hsp90. In cell lines selected for resistance to ganetespib, we have previously reported large chromosomal alterations (13). It was hypothesized that response rate to ganetespib triplet therapy might therefore be higher for patients harboring MPMs with a lower somatic copy number alterations. Genome-wide data was acquired for 11 of 27 patients [3 on 100 mg/m² ganetespib (2 cisplatin, 1 carboplatin), 8 on 200 mg/m² ganetespib (3 cisplatin, 5 carboplatin); remaining 16 patients provided non-viable samples]. Supplementary Table S4 shows the distribution of these variables over key baseline characteristics. As exploratory analyses, correlations and HRs between genomic variables and the change in total tumor burden and time to progression respectively are shown in Table 3, with mean and SEs of each genomic variable by best response category presented in Supplementary Table S5. Four patients did not have evaluable responses postbaseline. One of 7 response-evaluable patients with genomic data recorded (male, cisplatin cohort, 100 mg/m² ganetespib) had nonepithelioid histology and stable disease as best response. All other response-evaluable patients in this exploratory analysis achieved partial response and had epithelioid histology. Total LOH was associated with tumor shrinkage (i.e., higher baseline LOH associated with smaller reduction in total tumor burden; Spearman correlation = −0.703, P = 0.078; Supplementary Fig. S7) and a 12% increase in the risk of disease progression (HR = 1.12; 95% CI, 1.02–1.24, P = 0.018).

In MESO-02, we investigated the safety, tolerability, and efficacy of intravenous ganetespib combined with standard pemetrexed and platinum therapy in patients with MPM. The MTD of ganetespib was found to be 200 mg/m² in both the cisplatin and carboplatin cohorts. The trial successfully passed the phase Ib stage. However, the manufacturer of ganetespib decided not to proceed to the randomized phase II study following a strategic review and mixed findings from other solid tumor trials of ganetespib.

Ganetespib was well tolerated. At the MTD found in MESO-02, 3 patients of 18 (17%) experienced a DLT. Overall, 5 patients (18.5%) withdrew from the study due to unacceptable/serious adverse events, all of which were grade 2–3. This phase Ib study was not powered to detect improvements in efficacy measures compared with previous work. However, our results were highly encouraging: of 15 patients evaluable for response at the MTD, 10 (67%) had a best response of PR and 5 (33%) had stable disease. Median PFS at the MTD was 6.3 months and median OS at the MTD was 16.7 months. One patient (male, 63 years old at registration, epithelioid histology, baseline ECOG PS of 0) treated at the MTD received maintenance treatment with no observed progression after 60 cycles. While these results were obtained from a small sample of patients with MPM compared with other studies, they indicate that this ganetespib–pemetrexed–platinum therapy combination may be worth investigating in a larger randomized double-blind placebo-controlled trial, with histology, performance status, previous lines of therapy, and baseline LOH as key stratification factors.

Prior to the setup of MESO-02, most trials used 200 mg/m² ganetespib as monotherapy, given weekly for 3 weeks over a 4-week cycle. The regimen used in our trial was considered appropriate given the addition of pemetrexed and platinum agents, and that patients could continue to have ganetespib as maintenance. The aim was to ensure the majority of patients would complete and tolerate at least one cycle of treatment (24 of 27 patients completed at least one full cycle). While our study did not explore the tolerability of higher doses of ganetespib when combined with pemetrexed and platinum therapy, higher doses of ganetespib may still be tolerable.

MESO-02 recruited a higher percentage of male patients and patients with performance status of at least 1 compared with previous MPM studies (5, 7–10), which led to 93% of patients in our trial with a poor baseline EORTC prognosis score. Despite the relatively high rate of nonepithelioid MPM in the treated cohort (22%) and the trend toward lower response in this histologic subtype, there was a significant overall response rate (52%) with 62% of patients with epithelioid disease responding to treatment. These response rates are among the highest reported for any combination treatment in advanced MPM, although the confidence intervals are expectedly wide due to the small number of patients. However, this suggests potential activity over and above that of standard chemotherapy. Six patients had nonepithelioid MPM; 1 patient received the MTD of ganetespib (200 mg/m²), and another achieved a PR as best response. The poorer response of patients with nonepithelioid MPM may be due to the majority of patients (5 out of 6) receiving doses lower than the MTD.

These data are consistent with preclinical evidence supporting a role for Hsp90 in mediating DNA repair including homologous recombination, which may underpin synergy of Hsp90 inhibitors with platinum drugs (25, 26). Furthermore, thymidylate synthetase, which has been shown to correlate with pemetrexed activity is a putative mediator of antifolate resistance, and is downregulated following by Hsp90 inhibition (27–29).

Acquisition of aneuploidy has been reported to be associated with resistance to Hsp90 inhibition (15). We therefore hypothesized that patients harburing genomic instability (reflected in LOH) might exhibit resistance to ganetespib. In our exploratory analyses, we observed a statistically significant effect on time to progression, with shorter time to progression for more genomically unstable MPMs. However, the results of this need to be interpreted with caution as increasing genomic instability per se may be negatively prognostic, and that we only had a relatively small number of cases for these analyses (30). Nevertheless, our results indicate that patients harboring MPM with high levels of LOH may fail to benefit from addition of ganetespib. LOH was associated with a worse clinical outcome, when considering both total tumor burden and time to progression. Our study was underpowered to detect any interaction between specific copy number alterations and sensitivity to Hsp90 inhibition.

Chemoimmunotherapy has transformed the first-line treatment of non–small cell lung cancer and is currently being developed in studies such as DREAM (NCT04334759), PrE505 (NCT02899195), IND227 (NCT02784171), and BEAT-meso (NCT03762018). Of note, the single-arm PrE505 phase II study had an overall survival of 21 months, significantly greater than expected with the standard of care (31) and a phase III trial, PrE0506/DREAM3R, is planned. However, combined immune checkpoint inhibition (ICI) with ipilimumab and nivolumab was recently announced as being superior to pemetrexed–platinum in the pivotal Checkmate 743 phase III trial (32). This could herald an imminent change of practice in the first-line setting, creating new second-line development opportunities for novel, non-ICI or ICI–chemotherapy combinations. Recent evidence suggests that Hsp90 inhibition, through its upregulation of interferon response genes, can enhance ICI therapy and pave the way for possible future ICI or chemo-immunotherapy combination studies (20). In conclusion, ganetespib can be safely administered to patients with MPM at 200 mg/m² when combined with pemetrexed and platinum-based chemotherapy. This novel triplet also shows a potential signal of activity, for which further evaluation of ganetespib or other Hsp90 inhibitors should be done in a larger randomized trial.

D.A. Fennell reports personal fees from Aldeyra (research grant, advisory board), grants and personal fees from Bayer (research grant, advisory board) and Boehringer Ingelheim (research grant, speaker fees, advisory board), grants, personal fees, and non-financial support from BMS (research grant, advisory board, in-kind research support) and AstraZeneca (research grant, speaker fee, in-kind support), personal fees and nonfinancial support from MSD (speaker fee, in-kind research support), non-financial support from Clovis Oncology (in-kind research support) and Eli Lilly (in-kind research support), grants, personal fees, and other from Roche (research grant, speaker fee, in-kind research support) and nonfinancial support from Atlas Biomed (in-kind research support) and Inventiva (advisory board) outside the submitted work. S. Danson reports grants from Cancer Research UK & the Department of Health (to support the Sheffield Experimental Cancer Medicine Centre) during the conduct of the study. P.J. Woll reports personal fees from Theradex outside the submitted work. M. Forster reports grants and personal fees from AstraZeneca (institutional grant and personal honoraria), Bristol Myers Squibb (institutional grant and personal honoraria), MSD (institutional grant and personal honoraria) and Merck (institutional grant and personal honoraria), grants from Boehringer Ingelheim (institutional grant), personal fees from Achilles, Bayer, Celgene, Guardant Health, Nanobiotix, Novartis, Oxford VacMedix, Pfizer, Pharmamar, Roche, and Takeda outside the submitted work. A. Hackshaw reports grants from Synta Pharma (Synta provided a grant to UCL for trial coordination) during the conduct of the study. G.M. Wheeler reports personal fees from AstraZeneca (honorarium for organizing nonpromotional workshop) and personal fees from Novametrics Consulting (honorarium for delivering lectures at workshop) outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

D.A. Fennell: Conceptualization, resources, data curation, supervision, funding acquisition, investigation, writing-review and editing. S. Danson: Conceptualization, resources, investigation, writing-review and editing. P.J. Woll: Resources, investigation, writing-review and editing. M. Forster: Conceptualization, resources, investigation, writing-review and editing. D. Talbot: Conceptualization, resources, investigation, writing-review and editing. J. Child: Data curation, project administration, writing-review and editing. L. Farrelly: Resources, data curation, funding acquisition, project administration, writing-review and editing. A. Sharkey: Resources, investigation, methodology, writing-review and editing. S. Busacca: Resources, investigation, methodology, writing-review and editing. Y. Ngai: Data curation, funding acquisition, project administration, writing-review and editing. A. Hackshaw: Conceptualization, resources, formal analysis, funding acquisition, project administration, writing-review and editing. G.M. Wheeler: Software, formal analysis, writing-original draft, writing-review and editing.

We are especially grateful to all the patients and their clinicians for participating in this study. M. Forster is supported by the UCL/UCLH NIHR Biomedical Research Centre and runs early phase studies in the NIHR UCLH Clinical Research Facility supported by the UCL ECMC. Ganetespib was made available by Madrigal Pharmaceuticals Inc. (formerly Synta Pharmaceuticals Corp.). Aldeyra Therapeutics now owns ganetespib. The MESO-02 trial was reviewed, endorsed, and funded by Cancer Research UK (A11402). The funder was not involved in: the collection, analysis, or interpretation of data; the writing of study reports; or the decision to submit this article for publication.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.