Patients with Biallelic BRCA1/2 Inactivation Respond to Olaparib Treatment Across Histologic Tumor Types

This article is featured in Highlights of This Issue, p. 6067

Homologous recombination repair (HRR) is a crucial DNA repair pathway, essential for the repair of DNA double-strand breaks (DSBs; ref. 1) that the genome is continuously subjected to (2). It allows for error-free restoration of DNA integrity and sequence, even when the genomic damage is extensive. The breast cancer susceptibility genes BRCA1 and BRCA2 are two of the most extensively studied tumor suppressor genes and are key players in the homologous recombination (HR) pathway (3). Deleterious alterations in BRCA1 or BRCA2, both germline (4–6) and somatic (7, 8), result in deficient HRR (dHRR; refs. 9, 10) and a high risk of developing cancer. dHRR due to biallelic loss of function (LoF) mutations in BRCA1 or BRCA2 is seen in 4.9% of patients with cancer across tumor types (11–13).

Tumor cells with dHRR can be specifically targeted by drugs inducing multiple DNA strand breaks. Inhibitors of PARP specifically target the weakness of dHRR tumor cells (14–16) by synthetic lethality (17, 18) leading to selective cytotoxicity and apoptosis.

Olaparib, an oral inhibitor of PARP1, is approved by the FDA and European Medicines Agency (EMA) for several indications, among which the maintenance treatment of ovarian, fallopian tube, and primary peritoneal cancer with germline or somatic BRCA mutations after response to first-line platinum-based chemotherapy and, regardless of BRCA status, for recurrent ovarian, fallopian tube, and primary peritoneal cancer after response to platinum-based chemotherapy. Olaparib was most recently approved as monotherapy for patients with metastatic castration-resistant prostate cancer with germline or somatic BRCA mutations (EMA) and mutations in other homologous repair deficiency (HRD) genes (FDA; refs. 19–21). Several other PARP inhibitors (PARPis) have been registered for the treatment of epithelial ovarian, fallopian tube, and primary peritoneal cancer (rucaparib, ref. 22; niraparib, ref. 23) and gBRCAm breast cancer (talazoparib; ref. 24).

The majority of the phase II to III clinical trials performed focused on efficacy of PARPi monotherapy in BRCA-associated cancer types, often only based on the presence of a germline BRCA mutation, and lacking detailed biomarker information such as confirmation of biallelic BRCA LoF in tumor tissue. Data on the effectivity of PARPis in patients with somatic BRCA mutations are scarce.

In the Drug Rediscovery Protocol (DRUP, NCT02925234; ref. 25) patients are being treated based on their tumor molecular profile with registered targeted treatments outside their labeled indications, systematically recording efficacy and safety data. Moreover, the DRUP creates opportunities for extensive biomarker analysis by performing whole-genome sequencing (WGS) on baseline tumor biopsies. Within DRUP, we initiated a cohort in which patients were treated with olaparib based on a germline or somatic BRCA1 or BRCA2 LoF genomic event. Patients with a malignancy for which olaparib was not available as standard treatment were considered for this cohort. We hypothesized that a PARPi may be an effective treatment option for patients with malignant tumors harboring BRCA12 LoF mutations, both germline and somatic, independent of histology.

Here, we show that using PARPis is a potentially effective treatment strategy for patients with complete LoF of BRCA1/2 in the DRUP cohort of 24 patients “Olaparib for tumors with a BRCA1/2 mutation.” The importance of WGS, performed on baseline biopsies, is demonstrated by the correlation between complete LoF of BRCA1/2 and clinical benefit from olaparib. WGS provides information on both germline and somatic mutations, and genomic mutational signatures, allowing for optimal patient selection using a single assay.

The DRUP is an ongoing prospective, multicenter, nonrandomized basket trial in which patients with advanced solid malignancies are being treated on the basis of their tumor molecular profile, with targeted- or immunotherapy outside their registered indications (25). The basket trial design allows for an unlimited number of parallel cohorts consisting of patients with the same histologic tumor type, molecular target (defined at gene level), and study drug. Patients enrolled in the histology-agnostic cohort “Olaparib for tumors with a BRCA1 or BRCA2 mutation” received olaparib tablets 300 mg twice daily (26) in 28-day cycles until occurrence of disease progression or intolerable side effects. Dose reductions were allowed up to a minimum dose of 200 mg twice daily. Patients were enrolled in nine out of the 32 DRUP-participating hospitals in the Netherlands, between September 2016 and October 2019. To date, accrual in other cohorts of the DRUP is still ongoing.

This study is registered with ClinicalTrials.gov, number NCT02925234.

Adult patients with advanced solid malignancies, for which standard treatment options were exhausted, and with no option for on-label or phase III study treatment with PARPis, were enrolled. Expansion of the reimbursed indications of olaparib during the course of the trial resulted in exclusion of patients with the new “on-label” histologies from that moment on. Patients with those histologies who were already enrolled in DRUP were not excluded, but continued treatment within the trial and were included in the efficacy analysis. Preenrollment, patients needed to have a pathogenic, inactivating BRCA1 or BRCA2 mutation or deletion confirmed in their tumor tissue, identified using any validated genetic test within the context of routine diagnostics or using WGS in the context of the Dutch CPCT-02 study (27). At the start of the trial, confirmation of biallelic LoF of BRCA was not a requirement for eligibility yet. During the course of the trial, literature emerged reporting on the importance of complete LoF for response to PARPis. Therefore, we added biallelic LoF of BRCA as a second requirement for eligibility in this cohort. In all submitted cases, the variant was reviewed by two independent clinical molecular biologists, assessing the actionability of the variant. Actionable variants were homozygous deletions and inactivating biallelic somatic mutations or inactivating germline mutations with LOH. They advised the study team on the driver likelihood, after which the decision to include the patient was made by the study team.

For this cohort in DRUP, the general DRUP inclusion and exclusion criteria applied (25). Additionally, patients were not eligible if they had previously been treated with a PARPi, if they were immunocompromised, or if they had features suggestive of myelodysplastic syndrome (MDS) or acute myeloid leukemia (AML). Patients were considered evaluable for the primary endpoint if at least one cycle of olaparib was completed. Nonevaluable patients were excluded for the efficacy analysis, but included in the safety analysis.

The study is conducted in accordance with the International Conference of Harmonization of Good Clinical Practice and the Declaration of Helsinki, and was approved by the independent ethics committee and by the institutional review boards in every participating hospital. Patients provided written informed consent upon enrollment.

The primary end point of this study is the clinical benefit rate (CBR), defined as confirmed complete or partial response or stable disease for 16 weeks or more, according to RECIST 1.1 and measured at least twice, at least 28 days apart in a particular cohort. Tumor response was reported by the local investigator in the electronic case record form (eCRF).

Tumor assessments were done at baseline and after every second treatment cycle. If patients were on treatment ≥6 months, tumor assessments were performed after every three cycles. Secondary endpoints include: objective response rate (ORR, defined as partial or complete response), duration of response, progression-free survival (PFS), overall survival (OS), and treatment related Common Terminology Criteria for Adverse Events (CTCAE) grade ≥ 3 adverse events. Exploratory endpoints include biomarker analysis on fresh frozen tumor biopsies.

Safety is assessed by documentation of serious and study treatment-related grade ≥3 adverse events according to CTCAE v.4.03, and followed up until 1 month after the last dose of study drug. Safety within the trial is monitored by an Independent Data Monitoring Committee (IDMC) who is blinded for response rates per cohort during accrual.

Cohorts are monitored using a Simon-like two-stage “admissible” monitoring plan (28, 29) to identify cohorts with evidence of activity. Clinical benefit (CB) of ≥30% is considered of sufficient clinical interest to warrant further study in a confirmatory expansion cohort (stage III within the DRUP; ref. 30). The cohorts are evaluated in a two-stage design, if there would be no patients with CB in the first eight participants in the cohort, the cohort would be closed. Otherwise, an additional 16 patients would be included in the cohort. Four or fewer patients with CB out of 24 would suggest a lack of (clinically meaningful) activity, whereas 5 or more patients with CB would suggest that further investigation of the drug in the tumor/variant cohort is warranted. The null hypothesis and alternative hypothesis to be tested are defined as CBR of 10% versus ≥ 30%. This monitoring rule has 85% power to reject the null hypothesis of a CBR of 10% when the true CBR is 30%, with a one-sided alpha error rate of 7.8%. Exact 95% confidence intervals (CIs) were calculated using the Clopper–Pearson method.

At baseline, a new fresh frozen tumor biopsy was obtained from each patient. Biopsies were harvested and collected by the participating hospitals and sent to the Hartwig Medical Foundation (HMF), together with a 10-mL blood sample to determine the background variation of the germline DNA of the patient. For WGS, a minimum tumor-cell percentage of 30% is required. A 6-μm section was collected for hematoxylin and eosin (H&E) staining and estimation of tumor cellularity by an experienced pathologist. If the sample tumor cellularity was ≥ 30% and the DNA yield was ≥ 300 ng, WGS was performed.

WGS data were analyzed using an optimized, high-quality bioinformatic pipeline (31), and per patient a summarizing report of all relevant findings was created, including information on tumor purity, ploidy, somatic variants, copy-number variations, mutational load, and more complex genomic features such as gene fusions, Catalogue of Somatic Mutations in Cancer (COSMIC) mutational signatures (32) and microsatellite (in)stability. A Classifier of Homologous Recombination Deficiency (CHORD) for pan-cancer HRD detection, as recently developed by HMF, was computed for each sample, hereafter referred to as “HRD score” (33). Biallelic status of point mutations and the driver likelihood were assessed as described in previously published work (31). All code and scripts used for analysis of the WGS data are available at GitHub (https://github.com/hartwigmedical/).

Before biomarker analysis was performed, all WGS samples (baseline study samples and preenrollment WGS samples) were reanalyzed using the most recent HMF bioinformatics pipeline, including computation of the HRD score for each sample. The investigators and an independent clinical molecular biologist reviewed the baseline biopsy WGS results and confirmed presence of the qualifying BRCA mutation, assessed biallelic status of BRCA LoF and explored other identified oncogenic-driver alterations. In cases where no baseline WGS data were available (i.e., failed sequencing due to low tumor cellularity), the call for biallelic or mono-allelic BRCA LoF was made based on the preenrollment molecular data. If preenrollment WGS data was available, an HRD score was computed from that sample. Recent reports show a high spatiotemporal preservation of genomic-driver alterations (34) which justifies this approach.

This investigator-initiated study receives funding from the Dutch Cancer Society (KWF), Barcode for Life and receives equal funding from a number of pharmaceutical companies, including AstraZeneca. WGS was performed free of charge at HMF. Study medication was made available free of charge by the manufacturer.

AstraZeneca had no role in the design or execution of the study and no influence on the study report.

Between September 2016 and November 2019, 68 patients with advanced cancer harboring a BRCA1 or BRCA2 alteration, who had exhausted standard treatment options, were submitted to the study team for evaluation for potential study participation in the cohort “Olaparib for tumors with a BRCA1 or BRCA2 mutation.” Forty-five patients were approved by the study team to be screened for treatment with olaparib, 18 patients were ineligible for study participation (Supplementary Fig. S1). Twenty-seven patients with nine tumor types were found eligible and started study treatment, of which the majority (41%, N = 11) had prostate cancer. Nineteen patients were included despite their current labeled indication (prostate, n = 11; breast n = 3; ovarian, n = 3; and pancreatic cancer, n = 2), because at the time of enrollment PARPi treatment was still off-label and not reimbursed for their tumor type. Patients had a median number of four prior lines of systemic treatment (Table 1; Supplementary Table S1). The regimens varied greatly due to the different tumor types enrolled. Fifteen of 27 patients were treated with a platinum-containing regimen (carboplatin, n = 11; oxaliplatin, n = 3; and cisplatin, n = 1). Seven patients who were previously platinum resistant had clinical benefit of olaparib treatment. Three patients were not evaluable for the primary endpoint according to our protocol definition of evaluability and were excluded in the efficacy analysis (2 had clinical progression and rapid deterioration (within 4 weeks) before finishing the first complete cycle, one patient suffered from intolerable side effects and stopped study treatment after 6 days). All 27 patients who received at least one dose of study medication were included in the safety analysis. Baseline characteristics are presented in Table 1. Twenty-four patients were evaluated in the efficacy analysis. From here on, only the results and characteristics of these 24 patients are described.

Seventeen patients were included based on a BRCA2 mutation, and seven patients had a BRCA1 mutation. In 14 of 24 patients, the BRCA alteration was discovered by WGS, performed as part of the Dutch CPCT-02 study (27). In five patients, the target was found using an NGS panel (single molecular inversion probe–based sequencing analysis and/or multiplex ligation-dependent probe amplification). Four patients were included based on a germline test only, and in one patient, a germline test combined with two functional HRD tests was performed. This patient with breast cancer had a germline mono-allelic BRCA2 c.9104A>C mutation that was classified as a variant of uncertain significance. Functional characterization of this variant using embryonic stem cell complementation showed 50% reduction in HR functionality (35). In addition, a Recombination Capacity (RECAP) test (36) showed negative RAD51 staining after ex vivo irradiation of the tumor tissue, which is highly suggestive of HRD. On the basis of these results, the study team granted a waiver to include the patient. Twelve patients had a germline BRCA mutation. Six of them also had a somatic event in BRCA, or LOH in tumor tissue, resulting in complete BRCA LoF. Twelve patients were included based only on somatic BRCA alterations. In six of them, complete LoF of BRCA was confirmed preenrollment or based on the baseline WGS data (Supplementary Table S1).

Baseline study biopsies were performed in 22 out of 24 patients. For two patients, a biopsy was not possible for medical reasons. Thirteen (59%) biopsies were successfully sequenced. Eight biopsies could not be sequenced due to a low tumor cellularity (<30%) and one was sequenced despite a tumor cellularity below the threshold, confirming the qualifying BRCA mutation, but HRD score and biallelic call could not be extracted (Supplementary Table S1).

From seven of 13 patients with successful baseline biopsy WGS, preenrollment WGS data were also available. Additionally, from eight patients with failed baseline study WGS, preenrollment WGS data were available, and from three patients no WGS data were available and information on biallelic status and HRD score from these patients could not be retrieved. Based on a consensus of findings from the preenrollment and the baseline study biopsies, 15 out of 24 patients had confirmed biallelic BRCA LoF and a high HRD score (Supplementary Table S1). In two patients with prostate cancer the call for biallelic loss could not be made due to low tumor purity, but in one of them, the high HRD score suggests complete LoF of BRCA2. In six other patients, baseline WGS showed a low HRD score and only mono-allelic loss (n = 4) or no BRCA variant at all (n = 2, 9%; Supplementary Table S1).

Fourteen of 24 patients (58%, 95% CI, 37%–78%) had CB upon treatment with olaparib. The objective response rate was 29% (7/24 patients), median time on treatment was 5.8 months (95% CI, 1.8–9.2 months). At data cutoff (November 5, 2020), one patient was still on treatment. The median PFS in this cohort was 7 months (95% CI, 2–8 months) and the median OS was 13 months (95% CI, 7–NA months; Fig. 1). CB was observed across tumor types, including non-BRCA histologies such as cholangiocarcinoma, and in patients with both germline and somatic BRCA alterations (Fig. 2; Supplementary Table S1). In the group of patients with CB, the median treatment duration was 9.1 months (95% CI, 8.4–NA months). The difference in outcome between biallelic LoF of BRCA1 and BRCA2 was not statistically significant (Fisher exact value 0.2445).

CB was predominantly observed in patients with tumors harboring a biallelic LoF alteration of BRCA1 or BRCA2, and with an HRD genomic signature, with few exceptions: one patient with prostate cancer had prolonged stable disease, while having no signs of genomic biallelic BRCA loss. The preenrollment molecular data showed a somatic BRCA2 mutation with 24% variant allele frequency (VAF), while in the baseline study biopsy WGS data, no evidence of a BRCA alteration or HRD was found. As indicated before, the most likely cause of this discordance is tumor heterogeneity. It is known that patients with BRCA-associated tumor types can benefit from PARPis even if the tumor has only mono-allelic BRCA LoF (13). Another possible explanation for the clinical benefit in this patient may be that the dominant tumor clone indeed had a BRCA2 alteration, in combination with a posttranslational silencing of BRCA2, resulting in functional HRD. CB was also observed in two patients whose details regarding biallelic LoF and HRD score were unknown. Both patients had BRCA-associated tumor types and were included based on a germline test only, with no WGS results available to confirm the target. None of the four patients with confirmed mono-allelic loss had CB. Of the 15 patients with confirmed biallelic BRCA1/2 LoF, 11 had CB (73%).

Seven patients in this cohort had non–BRCA-associated tumor types. Of these, four (57%) had clinical benefit: two patients with cholangiocarcinoma, one with renal cell carcinoma, and one with endometrial cancer. WGS data showed biallelic LoF of BRCA (Supplementary Table S1). Three patients with non–BRCA-associated histologies had no benefit from olaparib. The WGS data from the two patients with colorectal cancer clearly showed no biallelic LoF of BRCA and no evidence of HRD. This suggests that the BRCA mutations found in these patients are likely neutral passenger mutations and a consequence rather than a cause of tumorigenesis, in line with previous reports (13). Both patients had TP53, APC, and KRAS mutations and one also had a SMAD4 mutation. One patient with adrenal gland carcinoma had biallelic BRCA LoF and HRD, however, a CTNNB1 (β-catenin) p.Ser45Pro mutation was found, suggestive for WNT pathway activation, which is a known mechanism of PARPi resistance via N⁶-methyladenosine modification of FZD10 mRNA, correlating with increased HR activity and reduced PARPi sensitivity (37). Additionally, this patient had a TP53 mutation and RB1 deletion (Supplementary Table S1).

Apart from the patient with adrenal gland carcinoma described above, three other patients had no CB, despite having BRCA-associated tumor types, confirmed biallelic BRCA LoF and a high HRD score. We analyzed WGS data to search for indicators of primary resistance to PARPis. In each patient, WGS analysis showed the presence of another (strong) oncogenic driver mutation that was not previously implied as possible PARPi resistance mechanism. One patient had breast cancer with an amplification (18 copies) of fibroblast growth factor receptor 1 (FGFR1), which is found in 6.9% to 19.7% of patients with metastatic breast cancer (31, 38) and has been reported as a possible driver alteration and potential therapeutic target in breast cancer (39–41). Another patient with pancreatic cancer had a homozygous loss of CDKN2A and a duplication of exons 3 to 6 of TGFBR2, likely leading to inactivation. CDKN2A (p16) is deleted or inactivated in 67% of patients with metastatic pancreatic cancer (31). If expressed, it compromises efficient BRCA1-dependent DNA repair (42) and it is associated with better radiosensitivity in vitro (43), while we hypothesize that the opposite may result in lower sensitivity to PARPis. Inactivation of TGFBR2 may also contribute to decreased sensitivity to PARPis because active TGFβ signaling in tumors enhances sensitivity to PARPis in vitro (44). In the third patient, also with pancreatic cancer, a KEAP p.Cys434* inactivating mutation, which is associated with drug resistance by regulation of expression of plasma membrane efflux pumps and detoxifying enzymes (45), and a KRAS p.Gly12Arg activating hotspot mutation were detected. In vitro cell line data have indicated a role of KRAS mutation for PARPi resistance (46), but the clinical relevance remains uncertain. In all these patients, it is likely that the tumors were not dependent on BRCA, but rather on another dominant non-HRD mutational process.

Serious adverse events (SAEs) occurred in 37% of the enrolled patients (Table 2). No unexpected toxicity or CTCAE grade ≥ 4 events were reported. Review of SAEs by the IDMC raised no safety concerns.

Precision medicine holds great promise for the future of patients with (advanced) cancer, but is hampered by many challenges, including target identification, prioritization and funding/reimbursement of biomarker identification and treatment, due to extremely low numbers of patients with similar molecular profiles. This makes established methods of randomized trials to generate solid evidence for determination of treatment benefit difficult. To circumvent this challenge, the innovative design of the DRUP allows evaluation of small groups of patients with rare cancer subtypes to determine the potential benefit of a targeted agent in a group of patients with a specific tumor molecular profile.

In patients with cancer harboring deleterious BRCA1/2 mutations, regardless of histologic tumor type, we here report that olaparib monotherapy is an effective and tolerable treatment option, for both germline and somatic alterations. The majority of patients (58%) derived CB from olaparib treatment. CB was almost exclusively observed in patients who had biallelic BRCA LoF and a high HRD score, confirming the absence of a functional homologous-repair system. Posthoc selection of only those patients with confirmed biallelic loss of BRCA1/2 (N = 15) revealed a CBR of 73% (N = 11).

A considerable proportion of patients in this cohort had BRCA-associated tumor types (i.e., prostate, ovarian, breast, and pancreatic cancer), of which we now know that olaparib is an effective treatment option (47–51). Ten out of 15 evaluable patients with BRCA-associated histologies had CB, which may in part contribute to the success of the cohort. Seven of 24 patients had non–BRCA-associated tumor types, of whom four (57%) had clinical benefit (Supplementary Table S1). These results indicate that patients with tumor types other than the known BRCA associated histologies can benefit from treatment with PARPis, provided that they have biallelic LoF of BRCA, resulting in HRD. It also emphasizes the importance of extensive molecular tumor profiling by means of WGS or large-panel sequencing for all patients. Small tumor-specific sequencing panels would, in all seven patients in this cohort, not have identified the BRCA mutations, as BRCA diagnostics is not part of the regular reimbursed care for these tumor types.

An important limitation of this study is the small sample size of 24 patients. Nine different tumor types were enrolled in this histology-agnostic cohort, resulting in a heterogeneous population with large variations in biological tumor features and previous treatments. The number of patients per tumor type is low, there is a relative underrepresentation of patients with non–BRCA-associated tumor types and since some important tumor types (i.e., non–small cell lung cancer) are not represented in our cohort, the results cannot simply be extrapolated to all patients with cancer. Though we find a clinically relevant signal of activity here, confirmation of our findings in a larger cohort is essential, with special emphasis on patients with non-BRCA tumor types.

Six patients ultimately did not have biallelic BRCA loss (mono-allelic loss: N = 4; no BRCA variant: N = 2). In two patients with prostate cancer, the qualifying BRCA variant could not be reidentified in the baseline biopsy WGS data, the exact reason for this discordance is unclear. No evidence for reversion of HRD (for example due to platinum-based chemotherapy) was found in the WGS data. A possible explanation in both cases could be inter or intratumor heterogeneity. Alternatively, in the first patient the low VAF of 24% may suggest that BRCA2 LoF was not the major driver of tumorigenesis and that the variant was lost in clonal evolution. However, the short time between preenrollment biopsy and baseline study biopsy did not support this. In the other patient with a BRCA1 exon-1 deletion, an alternative explanation could be that the deletion of exon 1, which is located outside the open reading frame and contains the BRCA1 promoter, could not be picked up by the bioinformatics pipeline. However, the low HRD score suggests that there was no functional HRD, which points toward the more likely hypothesis of tumor heterogeneity. In three other patients, the information regarding biallelic status of BRCA could not be retrieved. In the early days of the trial, confirmation of complete LoF of BRCA was not a requirement for eligibility. The initial inclusion of patients without complete LoF of BRCA1 or BRCA2 in this cohort may be considered a weakness but we regard it as an unintentional strength, as it underlines the importance of a sharply defined biomarker. Our data illustrate the contrast between the groups with and without complete LoF, in terms of CB to PARPi treatment (73% versus 17%). Clearly, this study is not powered to demonstrate a significant difference between these subgroups within the cohort due to the small number of patients. However, we noted this as an interesting signal that warrants confirmation in a larger independent cohort. Currently, pathologists and molecular biologists struggle to reliably call loss-of-heterogeneity and biallelic BRCA LoF using the available standard large NGS panels. Experts are able to circumvent some of the struggles by adding a custom design of polymorphous single-nucleotid polymorphisms (SNPs) around the BRCA1/2 genes, but this requires experience and expertise that is not widely available yet, and an uncertainty margin remains when NGS panels are used, especially for samples with lower tumor percentages. Due to the reliable detection of tumor purities, WGS facilitates the diagnostic process by accurately informing physicians on tumor-specific biallelic LoF of BRCA1/2 and HRD, as well as on the presence of additional mutations potentially causing resistance to PARPis, using a single assay. Prompt availability of this information allows for better patient selection for treatment with PARPis, preventing overtreatment of patients who will likely not benefit.

The availability of WGS data also allowed to explore possible reasons for unexplained lack of clinical benefit upon PARPi treatment in patients with HRD and biallelic BRCA LoF. As described, in the four patients who had no CB despite having a favorable HRD molecular profile, another dominant non-HRD mutational process was identified as possible explanation for the lack of benefit. Pan-cancer, it is known that tumors have a mean number of 5.7 candidate genomic driver events per patient (31), likely occurring at different stages of tumor evolution. Some tumors may have multiple drivers occurring as early events in tumor development. In tumors with HRD, not responding to PARPis, one could also hypothesize that biallelic BRCA LoF and HRD may simply manifest as a consequence of genomic instability rather than being an early driving genomic event, especially in late-stage cancers such as in our cohort. Although we did find potential underlying tumor biology contributing to resistance in these patients, it is still hypothetical and needs further investigation.

Although an association between clinical benefit from olaparib and platinum sensitivity has been described (52, 53), we here found that platinum-refractory tumors can still respond to PARPi treatment. Seven out of 12 patients previously treated with platinum-containing chemotherapy had CB upon olaparib treatment; one patient was primary resistant to carboplatin, which indicates that platinum-sensitivity alone may not be a good predictive biomarker for olaparib treatment outcome.

Baseline WGS was successfully performed on all biopsies that had sufficient tumor cellularity (N = 15; 60%). This is consistent with the overall WGS success rate within DRUP (25) and within the Dutch CPCT-02 study (31). Currently, the minimum required tumor cellularity for clinical-grade WGS analysis has been further downscaled from 30% to 20% due to ongoing technical improvements and optimized data analysis (bioinformatics; ref. 54), resulting in a current successful analysis of 71% (55).

The CBR observed in this cohort needs confirmation in a larger independent cohort. Currently, we are preparing an expansion cohort within DRUP. After the first example of a third-stage cohort “nivolumab for MSI tumors,” which is the first pilot of the new Dutch personalized reimbursement model that has been previously described (30), negotiations with the manufacturer, payers, and health authorities are currently ongoing to work toward a second expansion cohort in DRUP to study olaparib in BRCA1/2-mutated tumors. On the basis of our current findings and previous reports (13), we have refined the qualifying biomarker to biallelic (somatic or germline) LoF of BRCA1 or BRCA2, and only off-label tumor types (non-BRCA histologies) will be eligible. In this expansion cohort, the financial risk will be shared between the manufacturer of olaparib and the insurance companies. For the first 16 weeks of treatment, the study drug is provided by the manufacturer. Upon confirmation of clinical benefit at 16 weeks, the subsequent treatment will be reimbursed by the health care insurance on an individual basis while efficacy and safety data collection continues to ultimately support expansion of the existing labeled indications of the drug.

Olaparib is an effective treatment option for patients with cancer harboring somatic and germline deleterious BRCA1/2 alterations regardless of tumor type, who exhausted other treatment options. The CBR in this cohort was 58%, and CB was predominantly observed in patients harboring tumors with biallelic LoF of BRCA and HRD. In patients with non–BRCA-associated tumor types, 57% had clinical benefit, suggesting PARPis as a promising treatment strategy and justifying a broad molecular diagnostic approach in all patients. In patients in this cohort who had complete LoF of BRCA and HRD in tumor tissue, but without clinical benefit of olaparib, another potential oncogenic driver was discovered by WGS. Further investigation and confirmation of this CBR in patients with non-BRCA histologies in an independent expansion cohort is warranted, and is currently in preparation within DRUP for patients with biallelic BRCA LoF.

J.M. van Berge Henegouwen reports personal fees from AstraZeneca outside the submitted work. W.W.J. de Leng reports grants from Roche, BMS, and Pfizer and personal fees from Janssen outside the submitted work. N. Mehra reports grants and personal fees from AstraZeneca during the conduct of the study; grants, personal fees, and other support from MSD and Astellas; grants and personal fees from Roche, Janssen-Cilag, Pfizer, and BMS; and personal fees from Bayer outside the submitted work. M. Labots reports other support from Bristol Myers Squibb outside the submitted work. E.E. Voest reports grants from AstraZeneca during the conduct of the study and grants from AstraZeneca outside the submitted work. No disclosures were reported by the other authors.

H. van der Wijngaart: Data curation, formal analysis, investigation, visualization, writing–original draft, project administration. L.R. Hoes: Data curation, investigation, project administration, writing–review and editing. J.M. van Berge Henegouwen: Data curation, investigation, project administration, writing–review and editing. D.L. van der Velden: Data curation, investigation, project administration, writing–review and editing. L.J. Zeverijn: Data curation, investigation, project administration, writing–review and editing. P. Roepman: Formal analysis, writing–review and editing. E. van Werkhoven: Data curation, software, formal analysis, visualization, methodology, writing–review and editing. W.W.J. de Leng: Formal analysis, writing–review and editing. A.M.L. Jansen: Formal analysis, writing-review and editing. N. Mehra: Resources, writing–review and editing. D.G.J. Robbrecht: Resources, writing–review and editing. M. Labots: Resources, writing–review and editing. D.J.A. de Groot: Resources, writing–review and editing. A. Hoeben: Resources, writing–review and editing. P. Hamberg: Resources, writing–review and editing. H. Gelderblom: Conceptualization, supervision, investigation, writing–review and editing. E.E. Voest: Conceptualization, supervision, funding acquisition, investigation, methodology, writing–review and editing. H.M.W. Verheul: Conceptualization, formal analysis, supervision, investigation, writing–original draft.

This Investigator Initiated study receives funding from the KWF (grant number 10014/2016–1), Barcode for Life and receives equal funding from a number of pharmaceutical companies, among which AstraZeneca. Whole-genome sequencing was performed free of charge at the Hartwig Medical Foundation. Study medication was made available free of charge by the manufacturer.

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.