MultiCOVID: a multi modal deep learning approach for COVID-19 diagnosis

The rapid spread of the severe acute respiratory syndrome coronavirus 2 led to a global overextension of healthcare. Both Chest X-rays (CXR) and blood test have been demonstrated to have predictive value on Coronavirus Disease 2019 (COVID-19) diagnosis on different prevalence scenarios. With the objective of improving and accelerating the diagnosis of COVID-19, a multi modal prediction algorithm (MultiCOVID) based on CXR and blood test was developed, to discriminate between COVID-19, Heart Failure and Non-COVID Pneumonia and healthy (Control) patients. This retrospective single-center study includes CXR and blood test obtained between January 2017 and May 2020. Multi modal prediction models were generated using opensource DL algorithms. Performance of the MultiCOVID algorithm was compared with interpretations from five experienced thoracic radiologists on 300 random test images using the McNemar-Bowker test. A total of 8578 samples from 6123 patients (mean age 66 ± 18 years of standard deviation, 3523 men) were evaluated across datasets. For the entire test set, the overall accuracy of MultiCOVID was 84%, with a mean AUC of 0.92 (0.89-0.94). For 300 random test images, overall accuracy of MultiCOVID was significantly higher (69.6%) compared with individual radiologists (range, 43.7-58.7%) and the consensus of all five radiologists (59.3%, P < .001). Overall, we have developed a multimodal deep learning algorithm, MultiCOVID, that discriminates among COVID-19, heart failure, non-COVID pneumonia and healthy patients using both CXR and blood test with a significantly better performance than experienced thoracic radiologists.

Comprehensive NGS profiling to enable detection of ALK gene rearrangements and MET amplifications in non-small cell lung cancer

Introduction

Next-generation sequencing (NGS) is currently widely used for biomarker studies and molecular profiling to identify concurrent alterations that can lead to the better characterization of a tumor's molecular landscape. However, further evaluation of technical aspects related to the detection of gene rearrangements and copy number alterations is warranted.

Methods

There were 12 ALK rearrangement-positive tumor specimens from patients with non-small cell lung cancer (NSCLC) previously detected via fluorescence in situ hybridization (FISH), immunohistochemistry (IHC), and an RNA-based NGS assay, and 26 MET high gene copy number (GCN) cases detected by FISH, selected for this retrospective study. All 38 pre-characterized cases were reassessed utilizing the PGDx™ elio™ tissue complete assay, a 505 gene targeted NGS panel, to evaluate concordance with these conventional diagnostic techniques.

Results

The detection of ALK rearrangements using the DNA-based NGS assay demonstrated excellent sensitivity with the added benefit of characterizing gene fusion partners and genomic breakpoints. MET copy number alterations were also detected; however, some discordances were observed likely attributed to differences in algorithm, reporting thresholds and gene copy number state. TMB was also assessed by the assay and correlated to the presence of NSCLC driver alterations and was found to be significantly lower in cases with NGS-confirmed canonical driver mutations compared with those without (p=0.0019).

Discussion

Overall, this study validates NGS as an accurate approach for detecting structural variants while also highlighting the need for further optimization to enable harmonization across methodologies for amplifications.

Evaluation of the host immune response assay SeptiCyte RAPID for potential triage of COVID-19 patients

Tools for the evaluation of COVID-19 severity would help clinicians with triage decisions, especially the decision whether to admit to ICU. The aim of this study was to evaluate SeptiCyte RAPID, a host immune response assay (Immunexpress, Seattle USA) as a triaging tool for COVID-19 patients requiring hospitalization and potentially ICU care. SeptiCyte RAPID employs a host gene expression signature consisting of the ratio of expression levels of two immune related mRNAs, PLA2G7 and PLAC8, measured from whole blood samples. Blood samples from 146 adult SARS-CoV-2 (+) patients were collected within 48 h of hospital admission in PAXgene blood RNA tubes at Hospital del Mar, Barcelona, Spain, between July 28th and December 1st, 2020. Data on demographics, vital signs, clinical chemistry parameters, radiology, interventions, and SeptiCyte RAPID were collected and analyzed with bioinformatics methods. The performance of SeptiCyte RAPID for COVID-19 severity assessment and ICU admission was evaluated, relative to the comparator of retrospective clinical assessment by the Hospital del Mar clinical care team. In conclusion, SeptiCyte RAPID was able to stratify COVID-19 cases according to clinical severity: critical vs. mild (AUC = 0.93, p < 0.0001), critical vs. moderate (AUC = 0.77, p = 0.002), severe vs. mild (AUC = 0.85, p = 0.0003), severe vs. moderate (AUC = 0.63, p = 0.05). This discrimination was significantly better (by AUC or p-value) than could be achieved by CRP, lactate, creatine, IL-6, or D-dimer. Some of the critical or severe cases had "early" blood draws (before ICU admission; n = 33). For these cases, when compared to moderate and mild cases not in ICU (n = 37), SeptiCyte RAPID had AUC = 0.78 (p = 0.00012). In conclusion, SeptiCyte RAPID was able to stratify COVID-19 cases according to clinical severity as defined by the WHO COVID-19 Clinical Management Living Guidance of January 25th, 2021. Measurements taken early (before a patient is considered for ICU admission) suggest that high SeptiScores could aid in predicting the need for later ICU admission.

Plasmatic KRAS Kinetics for the Prediction of Treatment Response and Progression in Patients With KRAS-mutant Lung Adenocarcinoma

INTRODUCTION

KRAS is the most common driver mutation in lung cancer. ctDNA-based assessment offers advantages over tumor as a minimally invasive method able to capture tumor heterogeneity. Monitoring KRAS mutational load in ctDNA may be useful in the management of the patients.

METHODS

Consecutive patients diagnosed with KRAS mutant lung adenocarcinoma in the tumor biopsy were included in this study. Plasma samples were obtained at different time points during the course of the disease. KRAS mutations in plasma were quantified using digital PCR and correlated with mutations in tumor and with radiological response and progression.

RESULTS

Two hundred and forty-five plasma samples from 56 patients were analyzed. The rate of detection of KRAS mutations in plasma in our previously characterized KRAS-mutant cases was 82% overall, reaching 96% in cases with more than 1 metastatic location. The dynamics of KRAS mutational load predicted response in 93% and progression in 63% of cases, 33 and 50 days respectively in advance of radiological evaluation. Progression-free survival for patients in whom ctDNA was not detectable in plasma after treatment initiation was significantly longer than for those in whom ctDNA remained detectable (7.7 versus 3.2 months; HR: 0.44, p=0.004).

CONCLUSIONS

The detection of KRAS mutations in ctDNA showed a good correlation with that in tumor biopsy and, in most cases, predicted tumor response and progression to chemotherapy in advance of radiographic evaluation. The liquid biopsies for ctDNA-based molecular analyses are a reliable tool for KRAS testing in clinical practice.

Concomitant genomic alterations in KRAS mutant advanced lung adenocarcinoma

OBJECTIVES

KRAS mutations are one of the most prevalent alterations in non-small cell lung cancer. However, patients with this driver alteration present heterogeneous clinical outcomes. In this study, we have explored the potential clinical impact of coexisting alterations in this subset of patients.

MATERIALS AND METHODS

Samples from a cohort of 69 lung adenocarcinoma patients homogenously treated with platinum doublet as first-line therapy were evaluated using targeted next generation sequencing (NGS). Mutations and copy number alterations were assessed in 37 advanced KRAS-mutant (KRASm) and in 32 KRAS wild-type (KRASwt).

RESULTS

TP53 was the most frequent additional alteration found in both cohorts. Interestingly, TP53 mutations were more frequent in KRASwt than in KRASm patients (80 % vs. 34 %; p <  0.05) as well as STK11 mutations (17 % vs 8 %, p=NS). FGFR3 mutations were only found concomitantly with KRASm (11 %). No genomic co-alteration had an impact on overall survival within the KRASm patients treated with chemotherapy.

CONCLUSIONS

KRAS mutated lung adenocarcinoma is a heterogeneous entity and comprehensive characterization of co-alterations using NGS may lead to more accurate patient stratification.

Serum cytokine levels as predictive biomarkers of benefit from ipilimumab in small cell lung cancer

Background. Immunotherapy has shown efficacy in small cell lung cancer (SCLC), but only a subset of patients benefits. Surrogate biomarkers are urgently needed. Our aim was to evaluate serum Th1, Th2, and proinflammatory cytokines in two cohorts of SCLC patients before and during treatment with chemotherapy with or without ipilimumab and to correlate them with survival. Patients and methods. Two cohorts of SCLC patients were studied: patients treated with chemotherapy (n = 47), and patients treated with chemotherapy plus ipilimumab (n = 37). Baseline, on-treatment and after-treatment serum samples were evaluated for the presence of IL-1beta, IL-2, IL-4, IL-5, IL-6, IL-8, IL-10, IFN-gamma, TNF-alpha, GM-CSF, and Mip-1alpha using a Luminex assay. Differential changes in cytokines between cohorts were analyzed. Associations between cytokine levels and their changes with overall survival were evaluated. Results. Patients treated with ipilimumab showed a global increase of all cytokines after treatment initiation. A high level of IL-8 at baseline was associated with worse prognosis regardless of treatment. Baseline increased IL-2 levels predicted sensitivity to ipilimumab, while high IL-6 and TNF-alpha predicted resistance. An on-treatment increase in IL-4 levels in patients treated with immune-chemotherapy was associated with a better overall survival. Conclusions. The addition of ipilimumab to standard chemotherapy in SCLC modulates the serum levels of cytokines. Baseline levels and their change over time relate to overall survival. Blood-based biomarkers are convenient for patients, and our results support prospective validation of cytokines as predictive biomarkers for ipilimumab in SCLC.

Abstract LB-143: Th1/Th2 and inflammatory cytokines as biomarkers of response to ipilimumab in small cell lung cancer (SCLC) patients

Ipilimumab appears to benefit a subset of patients with SCLC but no predictive marker is available to date. Cytokines encompass a group of pleiotropic proteins secreted by immune and cancer cells with a crucial role in immune response (Th1/Th2 balance). Our study sought to evaluate the role of serum cytokine levels and changes during treatment as biomarkers of outcome with ipilimumab in SCLC.

Serial serum samples (pre and on-treatment at 9 weeks) from two cohorts (C) of patients with recently diagnosed SCLC were evaluated in this study. C1, with 47 patients treated with first-line platinum/etoposide (PE) and C2, with 37 patients treated with PE plus ipilimumab (10mg/kg) within the ICE trial. By means of a fluorescent multiplex immunobead assay (Luminex Technology) we analyzed serum concentrations of Th1 (IL-2, IFN-γ), Th2 (IL-4, -5, -6, -10) and inflammatory (IL-1β, IL-8, GM-CSF, TNF-α, MIP1-α) cytokines. Optimal cut-offs were determined by the Contal and O'Quigley method. We evaluated the associations of cytokine baseline levels and on-treatment changes with patient outcomes and compare these between cohorts. Statistical analysis was carried out with Stata/MP 14 (StataCorp LLC) and Prism 6.0 (GraphPad).

In the ipilimumab cohort, patients with increased baseline levels of IL-2 (&gt;1.67 pg/mL) showed a median overall survival (OS) of 24.2 months (m) (9.5-not reached) vs. those with lower levels [7.9m (5.6-18.46); p=0.030]. On the contrary, those with higher levels of IL-6 (&gt;3.74 pg/mL) or TNF-α (&gt;9.17 pg/mL) had worse OS (9.5m vs. 18.5m; p=0.019 and 7.8 vs. 18.5m; p=0.003, respectively). Lastly, patients whose IL-8 levels decreased from baseline to the on-treatment time-point (ICE) by more than 28.3%, had a worse OS [7.9m (3.9-17) vs. 18.46m (9.5-30.5); p=0.008]. The rest of associations observed were common between both cohorts.

Serum levels of Th1/Th2 and inflammatory cytokines change during treatments in SCLC patients. Having a ‘control' cohort of SCLC patients treated with chemotherapy allowed us to specifically identify differential changes induced by ipilimumab. Our study suggests that cytokines could be explored as biomarkers of sensitivity (IL-2) or resistance (IL-6 and TNF-α) to ipilimumab when combined with chemotherapy. Changes of cytokine levels during treatment might also predict benefit (IL-8). These findings warrant prospective validation.

Citation Format: Max Hardy-Werbin, Pedro Rocha, Oriol Arpí, Álvaro Taus, Xavier Durán Jordà, Deborah Joseph-Pietras, Ana Rovira, Joan Albanell, Christian Ottensmeier, Edurne Arriola. Th1/Th2 and inflammatory cytokines as biomarkers of response to ipilimumab in small cell lung cancer (SCLC) patients [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr LB-143.

Delay in starting radiotherapy due to neoadjuvant therapy does not worsen survival in unresected glioblastoma patients

PURPOSE

We retrospectively examined the potential effect on overall survival (OS) of delaying radiotherapy to administer neoadjuvant therapy in unresected glioblastoma patients.

PATIENTS AND METHODS

We compared OS in 119 patients receiving neoadjuvant therapy followed by standard treatment (NA group) and 96 patients receiving standard treatment without neoadjuvant therapy (NoNA group). The MaxStat package of R identified the optimal cut-off point for waiting time to radiotherapy.

RESULTS

OS was similar in the NA and NoNA groups. Median waiting time to radiotherapy after surgery was 13 weeks for the NA group and 4.2 weeks for the NoNA group. The longest OS was attained by patients who started radiotherapy after 12 weeks and the shortest by patients who started radiotherapy within 4 weeks (12.3 vs 6.6 months) (P = 0.05). OS was 6.6 months for patients who started radiotherapy before the optimal cutoff of 6.43 weeks and 19.1 months for those who started after this time (P = 0.005). Patients who completed radiotherapy had longer OS than those who did not, in all 215 patients and in the NA and NoNA groups (P = 0.000). In several multivariate analyses, completing radiotherapy was a universally favorable prognostic factor, while neoadjuvant therapy was never identified as a negative prognostic factor.

CONCLUSION

In our series of unresected patients receiving neoadjuvant treatment, in spite of the delay in starting radiotherapy, OS was not inferior to that of a similar group of patients with no delay in starting radiotherapy.

Dynamics of EGFR Mutation Load in Plasma for Prediction of Treatment Response and Disease Progression in Patients With EGFR-Mutant Lung Adenocarcinoma

BACKGROUND

The assessment of epidermal growth factor receptor (EGFR) mutations is crucial for the management of patients with lung adenocarcinoma. Circulating tumor DNA (ctDNA)-based assessment offers advantages over tumor as a minimally invasive method able to capture tumor heterogeneity.

PATIENTS AND METHODS

Consecutive patients diagnosed with EGFR-mutant lung adenocarcinoma in tumor biopsy were included in this study. Plasma samples were obtained at different time points during the course of the disease. EGFR mutations in plasma were quantified using BEAMing (beads, emulsions, amplification, and magnetics) or digital PCR and were correlated with mutations in tumor and with radiologic response and progression.

RESULTS

Two hundred twenty-one plasma samples from 33 patients were analyzed. EGFR mutations in plasma were detected in 83% of all patients and 100% of those with extrathoracic metastases. The dynamics of the EGFR mutation load predicted response in 93% and progression in 89% of cases well in advance of radiologic evaluation. Progression-free survival for patients in whom ctDNA was not detected in plasma during treatment was significantly longer than for those in whom ctDNA remained detectable (295 vs. 55 days; hazard ratio, 17.1; P < .001).

CONCLUSION

The detection of EGFR mutations in ctDNA showed good correlation with that in tumor biopsy and predicted tumor response and progression in most patients. The liquid biopsy for ctDNA-based assessment of EGFR mutations is a reliable technique for diagnosis and follow-up in patients with EGFR-mutant lung adenocarcinoma in routine clinical practice.

Assessment of neuronal autoantibodies in patients with small cell lung cancer treated with chemotherapy with or without ipilimumab

Small-cell lung cancer (SCLC) is often associated with paraneoplastic syndromes. To assess the role of anti-neuronal autoantibodies (NAAs) as biomarkers of treatment outcome, we assessed NAAs in serial samples from SCLC patients treated with chemoimmunotherapy compared to chemotherapy alone. We evaluated 2 cohorts: in cohort 1 (C1), 47 patients received standard platinum/etoposide, and in cohort 2 (C2), 38 patients received ipilimumab, carboplatin and etoposide. Serum samples at baseline and subsequent time points were analyzed for the presence of NAAs. NAAs were detected at baseline in 25 patients (53.2%) in C1 and in 20 patients (52.6%) in C2 (most frequently anti-Sox1). NAA at baseline was associated with limited disease (75% vs 50%; p: 0.096) and better overall survival (15.1 m vs 11.7 m; p: 0.032) in C1. Thirteen patients (28.9%) showed 2 or more reactivities before treatment; this was associated with worse PFS (5.5 m vs 7.3 m; p: 0.005) in patients treated with chemoimmunotherapy. NAA titers decreased after therapy in 68.9% patients, with no differential patterns of change between cohorts. Patients whose NAA titer decreased after treatment, showed longer OS [18.5 m (95% CI: 15.8 - 21.2)] compared with those whose NAA increased [12.3 m (95% CI: 8.1 - 16.5; p 0.049)], suggesting that antibody levels correlate to tumor load. Our findings reinforce the role of NAAs as prognostic markers and tumor activity/burden in SCLC, warrant further investigation in their predictive role for immunotherapy and raise concern over the use of immunotherapy in patients with more than one anti-NAA reactivity.

Heterogeneity of Tumor and Immune Cell PD-L1 Expression and Lymphocyte Counts in Surgical NSCLC Samples

BACKGROUND

Immune-checkpoint inhibitors against programmed cell death protein 1 (PD-1)/programmed death-ligand 1 (PD-L1) have shown remarkable therapeutic activity in non-small-cell lung cancer (NSCLC). However, biomarker-based patient selection remains a challenge. Our aim was to assess the heterogeneity of various immune markers between different tumor areas of surgically resected NSCLC specimens.

MATERIALS AND METHODS

We included 94 adenocarcinoma (ADC) and 50 squamous cell carcinoma (SCC) specimens. Two distinct tumor areas of each tumor sample were selected and incorporated into tissue microarrays. PD-L1 expression in tumor cells (TCs) and immune cells (ICs) was assessed using clone SP142 (Ventana). PDL1 gene amplification was assessed using fluorescence in situ hybridization. CD3 and CD8 densities were quantified using digital image-based analysis. Heterogeneity was assessed using kappa agreement index (KI) and intraclass correlation coefficient.

RESULTS

Prevalence of PD-L1 expression was 16.8% in TCs and 27.8% in ICs. Eleven tumors (7.6%) showed PDL1 amplification. In ADC, KI of PD-L1 TC and IC expression between cores was 0.465 and 0.260, compared with 0.274 and 0.124 in SCC, respectively. Higher concordance was observed for PDL1 amplification (KI, 0.647 in ADC and KI, 1 in SCC). Eleven (61.1%) of 18 amplified cores showed PD-L1 staining in < 5% of TCs. Intraclass correlation coefficients for CD3 and CD8 were 0.293 and 0.186 in ADC and 0.489 and 0.610 in SCC samples, respectively.

CONCLUSIONS

We found significant heterogeneity of PD-L1 expression in both ADC and SCC samples, especially in the IC compartment. Heterogeneous expression of PD-L1 could misclassify patients for PD-1/PD-L1-directed therapies.