Recalibration of a Deep Learning Model for Low-Dose Computed Tomographic Images to Inform Lung Cancer Screening Intervals.

Importance: Annual low-dose computed tomographic (LDCT) screening reduces lung cancer mortality, but harms could be reduced and cost-effectiveness improved by reusing the LDCT image in conjunction with deep learning or statistical models to identify low-risk individuals for biennial screening. Objective: To identify low-risk individuals in the National Lung Screening Trial (NLST) and estimate, had they been assigned a biennial screening, how many lung cancers would have been delayed 1 year in diagnosis. Design, Setting, and Participants: This diagnostic study included participants with a presumed nonmalignant lung nodule in the NLST between January 1, 2002, and December 31, 2004, with follow-up completed on December 31, 2009. Data were analyzed for this study from September 11, 2019, to March 15, 2022. Exposures: An externally validated deep learning algorithm that predicts malignancy in current lung nodules using LDCT images (Lung Cancer Prediction Convolutional Neural Network [LCP-CNN]; Optellum Ltd) was recalibrated to predict 1-year lung cancer detection by LDCT for presumed nonmalignant nodules. Individuals with presumed nonmalignant lung nodules were hypothetically assigned annual vs biennial screening based on the recalibrated LCP-CNN model, Lung Cancer Risk Assessment Tool (LCRAT + CT [a statistical model combining individual risk factors and LDCT image features]), and the American College of Radiology recommendations for lung nodules, version 1.1 (Lung-RADS). Main Outcomes and Measures: Primary outcomes included model prediction performance, the absolute risk of a 1-year delay in cancer diagnosis, and the proportion of people without lung cancer assigned a biennial screening interval vs the proportion of cancer diagnoses delayed. Results: The study included 10 831 LDCT images from patients with presumed nonmalignant lung nodules (58.7% men; mean [SD] age, 61.9 [5.0] years), of whom 195 were diagnosed with lung cancer from the subsequent screen. The recalibrated LCP-CNN had substantially higher area under the curve (0.87) than LCRAT + CT (0.79) or Lung-RADS (0.69) to predict 1-year lung cancer risk (P < .001). If 66% of screens with nodules were assigned to biennial screening, the absolute risk of a 1-year delay in cancer diagnosis would have been lower for recalibrated LCP-CNN (0.28%) than LCRAT + CT (0.60%; P = .001) or Lung-RADS (0.97%; P < .001). To delay only 10% of cancer diagnoses at 1 year, more people would have been safely assigned biennial screening under LCP-CNN than LCRAT + CT (66.4% vs 40.3%; P < .001). Conclusions and Relevance: In this diagnostic study evaluating models of lung cancer risk, a recalibrated deep learning algorithm was most predictive of 1-year lung cancer risk and had least risk of 1-year delay in cancer diagnosis among people assigned biennial screening. Deep learning algorithms could prioritize people for workup of suspicious nodules and decrease screening intensity for people with low-risk nodules, which may be vital for implementation in health care systems.

SARS-CoV-2 Genome-Based Severity Predictions Correspond to Lower qPCR Values and Higher Viral Load.

The 2019 coronavirus disease (COVID-19) pandemic has demonstrated the importance of predicting, identifying, and tracking mutations throughout a pandemic event. As the COVID-19 global pandemic surpassed one year, several variants had emerged resulting in increased severity and transmissibility. Here, we used PCR as a surrogate for viral load and consequent severity to evaluate the real-world capabilities of a genome-based clinical severity predictive algorithm. Using a previously published algorithm, we compared the viral genome-based severity predictions to clinically derived PCR-based viral load of 716 viral genomes. For those samples predicted to be "severe" (probability of severe illness >0.5), we observed an average cycle threshold (Ct) of 18.3, whereas those in in the "mild" category (severity probability <0.5) had an average Ct of 20.4 (P=0.0017). We also found a nontrivial correlation between predicted severity probability and cycle threshold (r = -0.199). Finally, when divided into severity probability quartiles, the group most likely to experience severe illness (≥75% probability) had a Ct of 16.6 (n = 10), whereas the group least likely to experience severe illness (<25% probability) had a Ct of 21.4 (n = 350) (P=0.0045). Taken together, our results suggest that the severity predicted by a genome-based algorithm can be related to clinical diagnostic tests and that relative severity may be inferred from diagnostic values.

Variants in SARS-CoV-2 associated with mild or severe outcome.

INTRODUCTION: The coronavirus disease 2019 (COVID-19) pandemic is a global public health emergency causing a disparate burden of death and disability around the world. The viral genetic variants associated with outcome severity are still being discovered. METHODS: We downloaded 155 958 severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) genomes from GISAID. Of these genomes, 3637 samples included useable metadata on patient outcomes. Using this subset, we evaluated whether SARS-CoV-2 viral genomic variants improved prediction of reported severity beyond age and region. First, we established whether including genomic variants as model features meaningfully increased the predictive power of our model. Next, we evaluated specific variants in order to determine the magnitude of association with severity and the frequency of these variants among SARS-CoV-2 genomes. RESULTS: Logistic regression models that included viral genomic variants outperformed other models (area under the curve = 0.91 as compared with 0.68 for age and gender alone; P < 0.001). We found 84 variants with odds ratios greater than 2 for outcome severity (17 and 67 for higher and lower severity, respectively). The median frequency of associated variants was 0.15% (interquartile range 0.09-0.45%). Altogether 85% of genomes had at least one variant associated with patient outcome. CONCLUSION: Numerous SARS-CoV-2 variants have 2-fold or greater association with odds of mild or severe outcome and collectively, these variants are common. In addition to comprehensive mitigation efforts, public health measures should be prioritized to control the more severe manifestations of COVID-19 and the transmission chains linked to these severe cases.Lay summary: This study explores which, if any, SARS-CoV-2 viral genomic variants are associated with mild or severe COVID-19 patient outcomes. Our results suggest that there are common genomic variants in SARS-CoV-2 that are more often associated with negative patient outcomes, which may impact downstream public health measures.

Using Prediction Models to Reduce Persistent Racial and Ethnic Disparities in the Draft 2020 USPSTF Lung Cancer Screening Guidelines.

We examined whether draft 2020 United States Preventive Services Task Force (USPSTF) lung cancer screening recommendations "partially ameliorate racial disparities in screening eligibility" compared with the 2013 guidelines, as claimed. Using data from the 2015 National Health Interview Survey, USPSTF-2020 increased eligibility by similar proportions for minorities (97.1%) and Whites (78.3%). Contrary to the intent of USPSTF-2020, the relative disparity (differences in percentages of model-estimated gainable life-years from National Lung Screening Trial-like screening by eligible Whites vs minorities) actually increased from USPSTF-2013 to USPSTF-2020 (African Americans: 48.3%-33.4% = 15.0% to 64.5%-48.5% = 16.0%; Asian Americans: 48.3%-35.6% = 12.7% to 64.5%-45.2% = 19.3%; Hispanic Americans: 48.3%-24.8% = 23.5% to 64.5%-37.0% = 27.5%). However, augmenting USPSTF-2020 with high-benefit individuals selected by the Life-Years From Screening with Computed Tomography (LYFS-CT) model nearly eliminated disparities for African Americans (76.8%-75.5% = 1.2%) and improved screening efficiency for Asian and Hispanic Americans, although disparities were reduced only slightly (Hispanic Americans) or unchanged (Asian Americans). The draft USPSTF-2020 guidelines increased the number of eligible minorities vs USPSTF-2013 but may inadvertently increase racial and ethnic disparities. LYFS-CT could reduce disparities in screening eligibility by identifying ineligible people with high predicted benefit regardless of race and ethnicity.

Pathogenic role of B-cell receptor signaling and canonical NF-κB activation in mantle cell lymphoma.

To interrogate signaling pathways activated in mantle cell lymphoma (MCL) in vivo, we contrasted gene expression profiles of 55 tumor samples isolated from blood and lymph nodes from 43 previously untreated patients with active disease. In addition to lymph nodes, MCL often involves blood, bone marrow, and spleen and is incurable for most patients. Recently, the Bruton tyrosine kinase (BTK) inhibitor ibrutinib demonstrated important clinical activity in MCL. However, the role of specific signaling pathways in the lymphomagenesis of MCL and the biologic basis for ibrutinib sensitivity of these tumors are unknown. Here, we demonstrate activation of B-cell receptor (BCR) and canonical NF-κB signaling specifically in MCL cells in the lymph node. Quantification of BCR signaling strength, reflected in the expression of BCR regulated genes, identified a subset of patients with inferior survival after cytotoxic therapy. Tumor proliferation was highest in the lymph node and correlated with the degree of BCR activation. A subset of leukemic tumors showed active BCR and NF-κB signaling apparently independent of microenvironmental support. In one of these samples, we identified a novel somatic mutation in RELA (E39Q). This sample was resistant to ibrutinib-mediated inhibition of NF-κB and apoptosis. In addition, we identified germ line variants in genes encoding regulators of the BCR and NF-κB pathway previously implicated in lymphomagenesis. In conclusion, BCR signaling, activated in the lymph node microenvironment in vivo, appears to promote tumor proliferation and survival and may explain the sensitivity of this lymphoma to BTK inhibitors.

Interactions between Ibrutinib and Anti-CD20 Antibodies: Competing Effects on the Outcome of Combination Therapy.

PURPOSE: Clinical trials of ibrutinib combined with anti-CD20 monoclonal antibodies (mAb) for chronic lymphocytic leukemia (CLL) report encouraging results. Paradoxically, in preclinical studies, in vitro ibrutinib was reported to decrease CD20 expression and inhibit cellular effector mechanisms. We therefore set out to investigate effects of in vivo ibrutinib treatment that could explain this paradox. EXPERIMENTAL DESIGN: Patients received single-agent ibrutinib (420 mg daily) on an investigator-initiated phase II trial. Serial blood samples were collected pretreatment and during treatment for ex vivo functional assays to examine the effects on CLL cell susceptibility to anti-CD20 mAbs. RESULTS: We demonstrate that CD20 expression on ibrutinib was rapidly and persistently downregulated (median reduction 74%, day 28, P < 0.001) compared with baseline. Concomitantly, CD20 mRNA was decreased concurrent with reduced NF-κB signaling. An NF-κB binding site in the promoter of MS4A1 (encoding CD20) and downregulation of CD20 by NF-κB inhibitors support a direct transcriptional effect. Ex vivo, tumor cells from patients on ibrutinib were less susceptible to anti-CD20 mAb-mediated complement-dependent cytotoxicity than pretreatment cells (median reduction 75%, P < 0.001); however, opsonization by the complement protein C3d, which targets cells for phagocytosis, was relatively maintained. Expression of decay-accelerating factor (CD55) decreased on ibrutinib, providing a likely mechanism for the preserved C3d opsonization. In addition, ibrutinib significantly inhibited trogocytosis, a major contributor to antigen loss and tumor escape during mAb therapy. CONCLUSIONS: Our data indicate that ibrutinib promotes both positive and negative interactions with anti-CD20 mAbs, suggesting that successfully harnessing maximal antitumor effects of such combinations requires further investigation.

Designing the furin-cleavable linker in recombinant immunotoxins based on Pseudomonas exotoxin A.

Recombinant immunotoxins (RITs) are fusion proteins that join antibodies to protein toxins for targeted cell killing. RITs armed with Pseudomonas exotoxin A (PE) are undergoing clinical trials for the treatment of cancer. The current design of PE-based RITs joins an antibody fragment to the catalytic domain of PE using a polypeptide linker that is cleaved by the protease furin. Intracellular cleavage of native PE by furin is required for cytotoxicity, yet the PE cleavage site has been shown to be a poor furin substrate. Here we describe the rational design of more efficiently cleaved furin linkers in PE-based RITs, and experiments evaluating their effects on cleavage and cytotoxicity. We found that changes to the furin site could greatly influence both cleavage and cytotoxicity, but the two parameters were not directly correlated. Furthermore, the effects of alterations to the furin linker were not universal. Identical mutations in the anti-CD22 RIT HA22-LR often displayed different cytotoxicity from mutations in the anti-mesothelin RIT SS1-LR/GGS, underscoring the prominent role of the target site in their intoxication pathways. Combining several beneficial mutations in HA22-LR resulted in a variant (HA22-LR/FUR) with a remarkably enhanced cleavage rate and improved cytotoxicity against five B cell lines and similar or enhanced cytotoxicity in five out of six hairy cell leukemia patient samples. This result informs the design of protease-sensitive linkers and suggests that HA22-LR/FUR may be a candidate for further preclinical development.

Health Disparities in the Immunoprevention of Human Papillomavirus Infection and Associated Malignancies.

Human papillomavirus (HPV) causes roughly 1.6% of the plus 1.6 million cases of cancer that are diagnosed in the United States each year. Despite the proven safety and efficacy of available vaccines, HPV remains the most common sexually transmitted infection. Underlying the high prevalence of HPV infection is the poor adherence to the Centers for Disease Control recommendation to vaccinate all 11- to 12-year-old males and females. In fact, only about 38 and 14% of eligible females and males, respectively, receive the complete, three-dose immunization. The many factors associated with missed HPV vaccination opportunities - including race, age, family income, and patient education - contribute to widespread disparities in vaccine completion and related health outcomes. Beyond patient circumstance, however, research indicates that the rigor and consistency of recommendation by primary care providers also plays a significant role in uptake of HPV immunization. Health disparities data are of vital importance to HPV vaccination campaigns because they can provide insight into how to address current problems and allocate limited resources where they are most needed. Furthermore, even modest gains in populations with low vaccination rates may yield great benefits because HPV immunization has been shown to provide herd immunity, indirect protection for non-immunized individuals achieved by limiting the spread of an infectious agent through a population. However, the impact of current HPV vaccination campaigns is hindered by stagnant immunization rates, which remain far below target levels despite a slow overall increase. Furthermore, gains in immunization are not equally distributed across gender, age, demographic, and socioeconomic divisions within the recommended group of vaccine recipients. To achieve the greatest impact, public health campaigns should focus on improving immunization coverage where it is weakest. They should also explore more subtle but potentially significant determinants of HPV vaccine initiation and completion, such as the attitudes of parents and healthcare providers and factors that exacerbate HPV-related health outcomes, including smoking and human immunodeficiency virus-mediated immunosuppression. Optimizing the efficacy of vaccination campaigns will require a health disparities approach that both identifies and remedies the underlying causes of population differences in HPV vaccination.

Harnessing the fcµ receptor for potent and selective cytotoxic therapy of chronic lymphocytic leukemia.

Chronic lymphocytic leukemia (CLL) is a B-cell malignancy in need of new, effective, and safe therapies. The recently identified IgM receptor FcµR is overexpressed on malignant B cells in CLL and mediates the rapid internalization and lysosomal shuttling of IgM via its Fc fragment (Fcµ). To exploit this internalization and trafficking pathway for targeted drug delivery, we engineered an IgM-derived protein scaffold (Fcµ) and linked it with the cytotoxic agent monomethylauristatin F. This Fcµ-drug conjugate was selectively toxic for FcµR-expressing cell lines in vitro and for CLL cells but not autologous normal T cells ex vivo. Notably, the cytotoxic activity of the Fcµ-drug conjugate was maintained in CLL cells carrying a 17p deletion, which predicts resistance to standard chemotherapy. Next, we tested the possible therapeutic application of the Fcµ-drug conjugate in immunodeficient NOD/SCID/IL-2Rγ(null) (NSG) mice engrafted with peripheral blood cells from patients with leukemia. Three intravenous injections of the Fcµ-drug conjugate over a 10-day period were well tolerated and selectively killed the human CLL cells but not the coengrafted autologous human T cells. In summary, we developed a novel strategy for targeted cytotoxic therapy of CLL based on the unique properties of FcµR. FcµR-targeted drug delivery showed potent and specific therapeutic activity in CLL, thus providing proof of concept for FcµR as a valuable therapeutic target in CLL and for IgM-based antibody-drug conjugates as a new targeting platform.