Lessons learned from postmarketing withdrawals of expedited approvals for oncology drug indications.

In the past decade, there have been a record number of oncology therapy approvals by the US Food and Drug Administration (FDA) and the European Medicines Agency (EMA). Besides the EMA's conditional marketing authorisation programme and the FDA's Accelerated Approval Program, we observe a tendency towards fast approval for exploratory studies with non-randomised, uncontrolled designs and surrogate endpoints. This issue raises concerns about the robustness and effectiveness of accepted treatments, leaving patients and health-care professionals in a state of uncertainty. A substantial number of accelerated approvals have recently been withdrawn in the USA, with some still authorised in Europe, emphasising discrepancies in regulatory standards that affect both patients and society as a whole. We highlight examples of drugs, authorised on the basis of surrogate endpoints, that were later withdrawn due to an absence of overall survival benefit. Our findings address the challenges and consequences of accelerated approval pathways in oncology. In conclusion, this Policy Review calls for regulatory bodies to better align their procedures and insist on robust evidence, preferably through unbiased randomised controlled trials. Drug approval processes should prioritise patient benefit, overall survival, and quality of life to minimise risks and uncertainties for patients.

A blood-based immune marker for resistance to pembrolizumab in patients with metastatic urothelial cancer.

PD1 inhibition is effective in patients with metastatic urothelial cancer (mUC), yet a large fraction of patients does not respond. In this study, we aimed to identify a blood-based immune marker associated with non-response to facilitate patient selection for anti-PD1. To this end, we quantified 18 immune cell populations using multiplex flow cytometry in blood samples from 71 patients with mUC (as part of a biomarker discovery trial; NCT03263039, registration date 28-08-2017). Patients were classified as responder (ongoing complete or partial response, or stable disease; n = 25) or non-responder (progressive disease; n = 46) according to RECIST v1.1 at 6 months of treatment with pembrolizumab. We observed no differences in numbers of lymphocytes, T-cells, granulocytes, monocytes or their subsets between responders and non-responders at baseline. In contrast, analysis of ratios of immune cell populations revealed that a high mature neutrophil-to-T-cell ratio (MNTR) exclusively identified non-responders. In addition, the survival of patients with high versus low MNTR was poor: median overall survival (OS) 2.2 vs 8.9 months (hazard ratio (HR) 6.6; p < 0.00001), and median progression-free survival (PFS) 1.5 vs 5.2 months (HR 5.6; p < 0.0001). The associations with therapy response, OS, and PFS for the MNTR were stronger than for the classical neutrophil-to-lymphocyte ratio (HR for OS 3.5, and PFS 3) and the PD-L1 combined positivity score (HR for OS 1.9, and PFS 2.1). In conclusion, the MNTR distinctly and uniquely identified non-responders to treatment and may represent a novel pre-treatment blood-based immune metric to select patients with mUC for treatment with pembrolizumab.

Is a History of Optimal Staging by Sentinel Lymph Node Biopsy in the Era Prior to Adjuvant Therapy Associated with Improved Outcome Once Melanoma Patients have Progressed to Advanced Disease?

INTRODUCTION: Sentinel lymph node biopsy (SLNB) is important for staging in patients with primary cutaneous melanoma. Did having previously undergone SLNB also affect outcomes in patients once they have progressed to metastatic melanoma in the era prior to adjuvant therapy? METHODS: Data were retrieved from the Dutch Melanoma Treatment Registry, a prospectively collected, nationwide database of patients with unresectable stage IIIC or IV (advanced) melanoma between 2012 and 2018. Melanoma-specific survival (MSS) was compared between patients with advanced cutaneous melanoma, previously treated with a wide local excision (WLE) or WLE combined with SLNB as initial treatment of their primary tumor. Cox regression analyses were used to analyze the influence of different variables on MSS. RESULTS: In total, 2581 patients were included, of whom 1412 were treated with a WLE of the primary tumor alone and 1169 in whom this was combined with SLNB. At a median follow-up of 44 months from diagnosis of advanced melanoma, MSS was significantly longer in patients who had previously undergone SLNB {median 23 months (95% confidence interval [CI] 19-29) vs. 18 months (95% CI 15-20) for patients treated with WLE alone; p = 0.002}. However, multivariate Cox regression did not identify SLNB as an independent favorable prognostic factor for MSS after diagnosis of advanced melanoma. CONCLUSION: Prior to the availability of adjuvant systemic therapy, once patients have unresectable stage IIIC or IV (advanced) melanoma, there was no difference in disease outcome for patients who were or were not previously staged with SLNB.

The PRO-RCC study: a long-term PROspective Renal Cell Carcinoma cohort in the Netherlands, providing an infrastructure for 'Trial within Cohorts' study designs.

BACKGROUND: Ongoing research in the field of both localized, locally advanced and metastatic renal cell carcinoma has resulted in the availability of multiple treatment options. Hence, many questions are still unanswered and await further research. A nationwide collaborative registry allows to collect corresponding data. For this purpose, the Dutch PROspective Renal Cell Carcinoma cohort (PRO-RCC) has been founded, for the prospective collection of long-term clinical data, patient reported outcome measures (PROMs) and patient reported experience measures (PREMs). METHODS: PRO-RCC is designed as a multicenter cohort for all Dutch patients with renal cell carcinoma (RCC). Recruitment will start in the Netherlands in 2023. Importantly, participants may also consent to participation in a 'Trial within cohorts' studies (TwiCs). The TwiCs design provides a method to perform (randomized) interventional studies within the registry. The clinical data collection is embedded in the Netherlands Cancer Registry (NCR). Next to the standardly available data on RCC, additional clinical data will be collected. PROMS entail Health-Related Quality of Life (HRQoL), symptom monitoring with optional ecological momentary assessment (EMA) of pain and fatigue, and optional return to work- and/or nutrition questionnaires. PREMS entail satisfaction with care. Both PROMS and PREMS are collected through the PROFILES registry and are accessible for the patient and the treating physician. TRIAL REGISTRATION: Ethical board approval has been obtained (2021_218) and the study has been registered at ClinicalTrials.gov (NCT05326620). DISCUSSION: PRO-RCC is a nationwide long-term cohort for the collection of real-world clinical data, PROMS and PREMS. By facilitating an infrastructure for the collection of prospective data on RCC, PRO-RCC will contribute to observational research in a real-world study population and prove effectiveness in daily clinical practice. The infrastructure of this cohort also enables that interventional studies can be conducted with the TwiCs design, without the disadvantages of classic RCTs such as slow patient accrual and risk of dropping out after randomization.

Impact of DNA damage repair defects and aggressive variant features on response to carboplatin-based chemotherapy in metastatic castration-resistant prostate cancer.

Platinum-based chemotherapy is not standard of care for unselected or genetically selected metastatic castration-resistant prostate cancer (mCRPC) patients. A retrospective assessment of 71 patients was performed on platinum use in the Netherlands. Genetically unselected patients yielded low response rates. For a predefined subanalysis of all patients with comprehensive next-generation sequencing, 30 patients were grouped based on the presence of pathogenic aberrations in genes associated with DNA damage repair (DDR) or aggressive variant prostate cancer (AVPC). Fourteen patients (47%) were DDR deficient (DDRd), of which seven with inactivated BRCA2 (BRCA2mut). Six patients classified as AVPC. DDRd patients showed beneficial biochemical response to carboplatin, largely driven by all BRCA2mut patients having >50% prostate-specific antigen (PSA) decline and objective radiographic response. In the wild-type BRCA2 subgroup, 35% had a >50% PSA decline (P = .006) and 16% radiographic response (P < .001). Median overall survival was 21 months for BRCA2mut patients vs 7 months (P = .041) for those with functional BRCA2. AVPC patients demonstrated comparable responses to non-AVPC, including a similar overall survival, despite the poor prognosis for this subgroup. In the scope of the registration of poly-(ADP)-ribose polymerase inhibitors (PARPi) for mCRPC, we provide initial insights on cross-resistance between PARPi and platinum compounds. By combining the literature and our study, we identified 18 patients who received both agents. In this cohort, only BRCA2mut patients treated with platinum first (n = 4), responded to both agents. We confirm that BRCA2 inactivation is associated with meaningful responses to carboplatin, suggesting a role for both PARPi and platinum-based chemotherapy in preselected mCRPC patients.

The effectiveness of monotherapy with PI3K/AKT/mTOR pathway inhibitors in ovarian cancer: A meta-analysis.

OBJECTIVE: To determine the clinical benefit of monotherapy with PI3K/AKT/mTOR inhibitors in patients diagnosed with advanced or recurrent ovarian cancer and to investigate the predictive value of current PI3K/AKT/mTOR biomarkers on therapy response. METHODS: A systematic search was conducted in PubMed, Embase and the Cochrane Library for articles reporting on treatment with PI3K/AKT/mTOR inhibitors in ovarian cancer. The primary endpoint was defined as the clinical benefit rate (CBR), including the proportion of patients with complete (CR) and partial response (PR) and stable disease (SD). Secondary endpoints included the overall response rate (ORR, including CR and PR) and drug-related grade 3 and 4 adverse events. RESULTS: We included 233 patients from 19 studies and observed a pooled CBR of 32% (95% CI 20-44%) and ORR of 3% (95% CI 0-6%) in advanced or recurrent ovarian cancer patients treated with PI3K/AKT/mTOR inhibitors. Subgroup analysis tended to favor the studies who selected patients based on current PI3K/AKT/mTOR biomarker criteria (e.g. genomic alterations or loss of PTEN protein expression), but the difference in CBR was not statistically significant from studies with unselected populations (respectively, CBR of 42% (95% CI 23-62%) and 27% (95% CI 14-42%), P = 0.217). To better reflect true patient benefit, we excluded SD <6 months as a beneficial outcome which resulted in a pooled CBR of 7% (95% CI 2-13%). The overall proportion of patients with drug-related grade 3 and 4 adverse events was 36%. CONCLUSIONS: The efficacy of monotherapy with PI3K/AKT/mTOR inhibitors in advanced recurrent ovarian cancer patients is limited to a small subgroup and selection of patients with the use of current biomarkers did not improved the CBR significantly. Given the toxicity profile, we suggest that current treatment with PI3K/AKT/mTOR inhibitors should not be initiated unless in clinical trials. Furthermore, improved biomarkers to measure functional PI3K/AKT/mTOR pathway activity are needed to optimize patient selection.

Validation of the updated renal graded prognostic assessment (GPA) for patients with renal cancer brain metastases treated with gamma knife radiosurgery.

INTRODUCTION: Prognosis of patients with brain metastasis (BM) from renal cell carcinoma (RCC) is relevant for treatment decisions and can be estimated with the Renal Graded Prognostic Assessment (GPA). The aim of this study is to validate the updated version of this instrument in a cohort treated with Gamma Knife radiosurgery (GKRS) without prior local intracerebral therapy. METHODS: Between 2007 and 2018, 106 RCC patients with BM were treated with GKRS. They were categorized according to the updated Renal GPA. Overall survival (OS), distant intracranial failure and local failure were estimated using the Kaplan-Meier method and risk factors were identified with Cox proportional hazard regressions. RESULTS: Median OS was 8.6 months. Median OS for GPA categories 0.0-1.0 (15%), 1.5-2.0 (12%), 2.5-3.0 (35%) and 3.5-4.0 (29%) was 2.9, 5.5, 8.1 and 20.4 months, respectively. Karnofsky performance status < 90, serum hemoglobin ≤ 12.5 g/dL, age > 65 years and time from primary diagnosis to brain metastasis < 1 year were significantly related with shorter survival, while presence of extracranial disease, the volume and total number of BM had no significant impact on OS. A total count of > 4 BM was the only predictive factor for distant intracranial failure, while none of the investigated factors predicted local failure. CONCLUSIONS: This study confirms the updated Renal GPA in an independent cohort as a valuable instrument to estimate survival in patients with BM from RCC treated with GKRS.

Real-world outcomes of radium-223 dichloride for metastatic castration resistant prostate cancer.

Aim: Timing of radium-223 (Ra-223) in metastatic castration-resistant prostate cancer (mCRPC) remains challenging due to alternative options and short window of opportunity. Methods: Ra-223 treated patients in the CAPRI-registry were included. Outcomes were evaluated based on treatment line of Ra-223. Results: Out of 285 patients, 49% received Ra-223 in line ≥3. 51% completed six Ra-223 injections and 34% had a symptomatic skeletal event after first Ra-223 without differences between subgroups. After correction of known prognostic factors Ra-223 in line ≥3 (HR: 3.267; 95% CI: 1.689-6.317; p < 0.01) remained associated with worse OS. Conclusion: In the Netherlands, Ra-223 was mainly started as second or third mCRPC-treatment in 2014-2018. Later timing of Ra-223 did affect OS, but not treatment completion and occurrence of symptomatic skeletal events.

Balancing the Optimal and the Feasible: A Practical Guide for Setting Up Patient Registries for the Collection of Real-World Data for Health Care Decision Making Based on Dutch Experiences.

OBJECTIVES: The aim of this article was to provide practical guidance in setting up patient registries to facilitate real-world data collection for health care decision making. METHODS: This guidance was based on our experiences and involvement in setting up patient registries in oncology in the Netherlands. All aspects were structured according to 1) mission and goals ("the Why"), 2) stakeholders and funding ("the Who"), 3) type and content ("the What"), and 4) identification and recruitment of patients, data handling, and pharmacovigilance ("the How"). RESULTS: The mission of most patient registries is improving patient health by improving the quality of patient care; monitoring and evaluating patient care is often the primary goal ("the Why"). It is important to align the objectives of the registry and agree on a clear and functional governance structure with all stakeholders ("the Who"). There is often a trade off between reliability, validity, and specificity of data elements and feasibility of data collection ("the What"). Patient privacy should be carefully protected, and address (inter-)national and local regulations. Patient registries can reveal unique safety information, but it can be challenging to comply with pharmacovigilance guidelines ("the How"). CONCLUSIONS: It is crucial to set up an efficient patient registry that serves its aims by collecting the right data of the right patient in the right way. It can be expected that patient registries will become the new standard alongside randomized controlled trials due to their unique value.

CAST: A retrospective analysis of cabazitaxel and abiraterone acetate sequential treatment in patients with metastatic castrate-resistant prostate cancer previously treated with docetaxel.

Cabazitaxel and abiraterone have both received approval for treating metastatic castrate-resistant prostate cancer (mCRPC) patients after first-line docetaxel therapy. In the cabazitaxel and abiraterone sequential treatment (CAST) study, the clinical outcome of docetaxel-treated mCRPC patients treated sequentially with both cabazitaxel and abiraterone was studied. Data were collected retrospectively from mCRPC patients at 12 hospitals across the Netherlands who initiated cabazitaxel and/or abiraterone before December 2012. Primary outcome measure was overall survival (OS); secondary measures were progression-free survival (PFS), biochemical PFS, and best clinical and PSA response. Hospital admission data during treatment were collected, as well as toxicities resulting in treatment discontinuation or patient death. Sixty-three and 69 patients received Cab→Abi (cabazitaxel prior to abiraterone) and Abi→Cab before July 10th, 2013, respectively. Median OS was 19.1 months and 17.0 months in Cab→Abi and Abi→Cab treated patients, respectively (p = 0.369). Median PFS and biochemical PFS were significantly longer in Cab→Abi treated patients: 8.1 versus 6.5 (p = 0.050) and 9.5 versus 7.7 months (p = 0.024), respectively. Although partial responses to cabazitaxel occurred in both groups, Abi→Cab treated patients had a significantly decreased antitumor response from cabazitaxel than Cab→Abi treated patients (median PFS 5.0 versus 2.6 months, p < 0.001). Minor differences in toxicities were observed based on therapy sequence; generally, toxicity from cabazitaxel could be severe, while abiraterone toxicity was milder. This retrospective analysis indicates that primary progression on cabazitaxel or abiraterone did not preclude a response to the other agent in mCRPC patients. However, tumor response of both agents, particularly cabazitaxel, was lower when administered as higher-line therapy in the selected study population.

Bronchiolitis after Combination Immunotherapy With Ipilimumab and Nivolumab in a Melanoma Patient.

Therapy with immune checkpoint inhibitors (ICIs) has significantly improved the prognosis of metastatic melanoma but is also associated with various immune-related adverse events (AE), including pulmonary toxicity. Herein, we describe the case of a 60-year-old female with metastasized melanoma with BRAF mutation under combination immunotherapy with ipilimumab and nivolumab, who presented with a persistent, nonproductive cough for the last two months. Her CT-scan showed de novo bronchial inflammation and wall thickening in all lung fields. Initial treatment with antimicrobial treatment and inhalation corticosteroids did not resolve her symptoms, nor the radiologic abnormalities. Additional testing with transbronchial cryobiopsy showed a histologic picture of diffuse ill-formed granulomas and the presence of moderate chronic active inflammation of the respiratory epithelium, consistent with medication-related bronchiolitis. Bronchiolitis, as present in this case, has rarely been reported as an immune-related AE. A thorough diagnostic workup is mandatory as it remains a diagnosis of exclusion. Management consists of discontinuing ICIs and administering systemic corticosteroids. The addition of immunosuppressive agents (e, infliximab, cyclophosphamide, or mycophenolate mofetil) can be considered in refractory cases. In our case, clinical and radiologic resolution was achieved after discontinuing the ICI and treatment with high-dose prednisone. This case shows that although bronchiolitis is a rare immune-related side effect of ICIs, oncologists, and pulmonologists should always be aware of this relatively easily treatable AE.

Phenotype-guided targeted therapy based on functional signal transduction pathway activity in recurrent ovarian cancer patients: The STAPOVER study protocol.

Objective: Ovarian cancer is the fifth cause of cancer-related death among women. The benefit of targeted therapy for ovarian cancer patients is limited even if treatment is stratified by molecular signature. There remains a high unmet need for alternative diagnostics that better predict targeted therapy, as current diagnostics are generally inaccurate predictors. Quantitative assessment of functional signal transduction pathway (STP) activity from mRNA measurements of target genes is an alternative approach. Therefore, we aim to identify aberrantly activated STPs in tumour tissue of patients with recurrent ovarian cancer and start phenotype-guided targeted therapy to improve survival without compromising quality of life. Study design: Patients with recurrent ovarian cancer and either 1) have platinum-resistant disease, 2) refrain from standard therapy or 3) are asymptomatic and not yet eligible for standard therapy will be included in this multi-centre prospective cohort study with multiple stepwise executed treatment arms. Targeted therapy will be available for patients with aberrantly high functional activity of the oestrogen receptor, androgen receptor, phosphoinositide 3-kinase or Hedgehog STP. The primary endpoint of this study is the progression-free survival (PFS) ratio (PFS2/PFS1 ratio) according to RECIST 1.1 determined by the PFS on matched targeted therapy (PFS2) compared to PFS on prior therapy (PFS1). Secondary endpoints include among others best overall response, overall survival, side effects, health-related quality of life and cost-effectiveness. Conclusion: The results of this study will show the clinical applicability of STP activity in selecting recurrent ovarian cancer patients for effective therapies.

Gene-expression-based T-Cell-to-Stroma Enrichment (TSE) score predicts response to immune checkpoint inhibitors in urothelial cancer.

Immune checkpoint inhibitors (ICI) improve overall survival in patients with metastatic urothelial cancer (mUC), but therapeutic success at the individual patient level varies significantly. Here we identify predictive markers of response, based on whole-genome DNA (n = 70) and RNA-sequencing (n = 41) of fresh metastatic biopsy samples, collected prior to treatment with pembrolizumab. We find that PD-L1 combined positivity score does not, whereas tumor mutational burden and APOBEC mutagenesis modestly predict response. In contrast, T cell-to-stroma enrichment (TSE) score, computed from gene expression signature data to capture the relative abundance of T cells and stromal cells, predicts response to immunotherapy with high accuracy. Patients with a positive and negative TSE score show progression free survival rates at 6 months of 67 and 0%, respectively. The abundance of T cells and stromal cells, as reflected by the TSE score is confirmed by immunofluorescence in tumor tissue, and its good performance in two independent ICI-treated cohorts of patients with mUC (IMvigor210) and muscle-invasive UC (ABACUS) validate the predictive power of the TSE score. In conclusion, the TSE score represents a clinically applicable metric that potentially supports the prospective selection of patients with mUC for ICI treatment.

Multi-omic analysis identifies hypoalbuminemia as independent biomarker of poor outcome upon PD-1 blockade in metastatic melanoma.

We evaluated the prognostic value of hypoalbuminemia in context of various biomarkers at baseline, including clinical, genomic, transcriptomic, and blood-based markers, in patients with metastatic melanoma treated with anti-PD-1 monotherapy or anti-PD-1/anti-CTLA-4 combination therapy (n = 178). An independent validation cohort (n = 79) was used to validate the performance of hypoalbuminemia compared to serum LDH (lactate dehydrogenase) levels. Pre-treatment hypoalbuminemia emerged as the strongest predictor of poor outcome for both OS (HR = 4.01, 95% CI 2.10-7.67, Cox P = 2.63e-05) and PFS (HR = 3.72, 95% CI 2.06-6.73, Cox P = 1.38e-05) in univariate analysis. In multivariate analysis, the association of hypoalbuminemia with PFS was independent of serum LDH, IFN-γ signature expression, TMB, age, ECOG PS, treatment line, treatment type (combination or monotherapy), brain and liver metastasis (HR = 2.76, 95% CI 1.24-6.13, Cox P = 0.0131). Our validation cohort confirmed the prognostic power of hypoalbuminemia for OS (HR = 1.98, 95% CI 1.16-3.38; Cox P = 0.0127) and was complementary to serum LDH in analyses for both OS (LDH-adjusted HR = 2.12, 95% CI 1.2-3.72, Cox P = 0.00925) and PFS (LDH-adjusted HR = 1.91, 95% CI 1.08-3.38, Cox P = 0.0261). In conclusion, pretreatment hypoalbuminemia was a powerful predictor of outcome in ICI in melanoma and showed remarkable complementarity to previously established biomarkers, including high LDH.

Efficacy of pembrolizumab and biomarker analysis in patients with WGS-based intermediate to high tumor mutational load: results from the Drug Rediscovery Protocol.

PURPOSE: To evaluate efficacy of pembrolizumab across multiple cancer types harboring different levels of Whole-Genome Sequencing (WGS)-based tumor mutational load (TML; total of non-synonymous mutations across the genome) in patients included in the Drug Rediscovery Protocol (NCT02925234). PATIENTS AND METHODS: Patients with solid, treatment-refractory, microsatellite-stable tumors were enrolled in cohort A: breast cancer TML 140-290, cohort B: tumor-agnostic cohort TML 140-290, and cohort C: tumor-agnostic cohort TML >290. Patients received pembrolizumab 200 mg every three weeks. Primary endpoint was clinical benefit (CB: objective response or stable disease (SD) ≥16 weeks). Pre-treatment tumor biopsies were obtained for WGS and RNA-sequencing. RESULTS: Seventy-two evaluable patients with 26 different histotypes were enrolled. CB rate was 13% in cohort A (3/24 with partial response (PR)), 21% in cohort B (3/24 with SD, 2/24 with PR), and 42% in cohort C (4/24 with SD, 6/24 with PR). In cohort C, neoantigen burden estimates and expression of inflammation and innate immune biomarkers were significantly associated with CB. Similar associations were not identified in cohort A and B. In cohort A, CB was significantly associated with mutations in the chromatin remodeling gene PBRM1, while in cohort B, CB was significantly associated with expression of MICA/MICB and butyrophilins. CB and clonal TML were not significantly associated. CONCLUSION: While in cohort A pembrolizumab lacked activity, cohort B and cohort C met the study's primary endpoint. Further research is warranted to refine selection of patients with tumors harboring lower TMLs and may benefit from a focus on innate immunity.

[Promosing cancer drug, but not available; the complexity of alternative access procedures].

It can be a challenge for a healthcare system to facilitate access to novel cancer drugs. This can be due to lack of transparency regarding alternative access pathways before or after authorisation, absence of reimbursement due to uncertainties regarding clinical value and cost-effectiveness, lengthy price-negotiations and practice variation among hospitals, for instance in conducting molecular diagnostics. Based on three daily practice examples, we discuss the current challenges and complex procedures in providing access through alternative pathways, such as early-access or free-of-charge programmes in the Netherlands. Policymakers, funders, and medical societies need to harmonise procedures, increase transparency regarding treatment and testing options and collaborate more to ensure equal and timely access to valuable cancer therapies. Accurate and responsible communication and reporting regarding the efficacy and availability of novel treatments is crucial to prevent unrealistic expectations, disappointment, and distress among patients and the public.

Incidence and survival of castration-resistant prostate cancer patients with visceral metastases: results from the Dutch CAPRI-registry.

BACKGROUND: The objective of this real-world population study is to investigate incidence and treatment of visceral metastases (VMs) in castration resistant prostate cancer (CRPC) patients and their survival. METHODS: CRPC-patients in the CAPRI-registry between 2010 and 2016 were included in the analyses and followed till 2017. Outcomes were proportion of patients radiologically screened for VMs and proportion of patients with VMs at CRPC-diagnosis and at the start of every treatment line. Groups have been created based on location of VMs (lung, liver, or both) at date of first VM diagnosis. The outcome for these groups was overall survival (OS). Statistics included descriptive analyses, Kaplan-Meier method, and Cox proportional hazard regression analysis for survival analyses. RESULTS: Of 3602 patients from the CAPRI registry, 457 patients (12.7%) were diagnosed with VMs during follow-up: 230 patients with liver, 161 with lung, and 66 with both liver and lung metastases. The proportion of patients radiologically screened for VMs increased per treatment line as did the occurrence rate of VMs. However, 80% of patients at CRPC diagnosis to 40% in the 6th line were not screened for VMs at the start of a systemic treatment. Median OS was 8.6 months for patients with liver, 18.3 with lung and 10.9 with both liver and lung metastases (p < 0.001) from date of first VM diagnosis. After correction for prognostic factors patients with lung metastases had significantly better OS than patients with liver metastases (HR 0.650, p = 0.001). CONCLUSION: This real-world analysis showed that despite the increased rate of radiological staging during follow-up, still 80% to 40% of the patients (CRPC diagnosis to 6th treatment line respectively) were not screened for VMs at the start of a systemic treatment. VMs and location of VMs are key prognostic patient characteristics, impacts survival and have implications for treatment decisions, so routine staging of CRPC-patients is warranted. CLINICAL TRIAL IDENTIFICATION: The CAPRI study is registered in the Dutch Trial Registry as NL3440 (NTR3591).

CT radiomics compared to a clinical model for predicting checkpoint inhibitor treatment outcomes in patients with advanced melanoma.

INTRODUCTION: Predicting checkpoint inhibitors treatment outcomes in melanoma is a relevant task, due to the unpredictable and potentially fatal toxicity and high costs for society. However, accurate biomarkers for treatment outcomes are lacking. Radiomics are a technique to quantitatively capture tumour characteristics on readily available computed tomography (CT) imaging. The purpose of this study was to investigate the added value of radiomics for predicting clinical benefit from checkpoint inhibitors in melanoma in a large, multicenter cohort. METHODS: Patients who received first-line anti-PD1±anti-CTLA4 treatment for advanced cutaneous melanoma were retrospectively identified from nine participating hospitals. For every patient, up to five representative lesions were segmented on baseline CT, and radiomics features were extracted. A machine learning pipeline was trained on the radiomics features to predict clinical benefit, defined as stable disease for more than 6 months or response per RECIST 1.1 criteria. This approach was evaluated using a leave-one-centre-out cross validation and compared to a model based on previously discovered clinical predictors. Lastly, a combination model was built on the radiomics and clinical model. RESULTS: A total of 620 patients were included, of which 59.2% experienced clinical benefit. The radiomics model achieved an area under the receiver operator characteristic curve (AUROC) of 0.607 [95% CI, 0.562-0.652], lower than that of the clinical model (AUROC=0.646 [95% CI, 0.600-0.692]). The combination model yielded no improvement over the clinical model in terms of discrimination (AUROC=0.636 [95% CI, 0.592-0.680]) or calibration. The output of the radiomics model was significantly correlated with three out of five input variables of the clinical model (p < 0.001). DISCUSSION: The radiomics model achieved a moderate predictive value of clinical benefit, which was statistically significant. However, a radiomics approach was unable to add value to a simpler clinical model, most likely due to the overlap in predictive information learned by both models. Future research should focus on the application of deep learning, spectral CT-derived radiomics, and a multimodal approach for accurately predicting benefit to checkpoint inhibitor treatment in advanced melanoma.