Autonomous artificial intelligence versus teleophthalmology for diabetic retinopathy.

Purpose: To assess the role of artificial intelligence (AI) based automated software for detection of Diabetic Retinopathy (DR) compared with the evaluation of digital retinography by two double masked retina specialists. Methods: Two-hundred one patients (mean age 65 ± 13 years) with type 1 diabetes mellitus or type 2 diabetes mellitus were included. All patients were undergoing a retinography and spectral domain optical coherence tomography (SD-OCT, DRI 3D OCT-2000, Topcon) of the macula. The retinal photographs were graded using two validated AI DR screening software (Eye Art TM and IDx-DR) designed to identify more than mild DR. Results: Retinal images of 201 patients were graded. DR (more than mild DR) was detected by the ophthalmologists in 38 (18.9%) patients and by the AI-algorithms in 36 patients (with 30 eyes diagnosed by both algorithms). Ungradable patients by the AI software were 13 (6.5%) and 16 (8%) for the Eye Art and IDx-DR, respectively. Both AI software strategies showed a high sensitivity and specificity for detecting any more than mild DR without showing any statistically significant difference between them. Conclusions: The comparison between the diagnosis provided by artificial intelligence based automated software and the reference clinical diagnosis showed that they can work at a level of sensitivity that is similar to that achieved by experts.

Pulmonary Sarcoidosis and Immune Dysregulation: A Pilot Study on Possible Correlation.

BACKGROUND: Sarcoidosis is a systemic inflammatory disease characterized by an altered inflammatory response. OBJECTIVE: The aim of this study was to evaluate whether immune system alterations detected by lymphocyte typing in peripheral blood correlate with the severity of sarcoidosis, calculated according to two separate severity scores proposed by Wasfi in 2006 and Hamzeh in 2010. MATERIALS AND METHODS: Eighty-one patients were recruited, and clinical data and laboratory tests at the time of diagnosis were obtained in order to assess the severity index score and investigate any statistically significant correlation with the cytofluorimetry data. RESULTS: Our data demonstrated that none of the two scores show an association with the level of total lymphocytes or lymphocyte subclasses. LIMITATIONS: First of all, the sample taken into consideration is small. The assessment was performed only at disease onset and not during the disease. Furthermore, the severity scores do not take into account disease activity (measured by PET/CT or gallium scintigraphy). CONCLUSIONS: Lymphocyte subpopulation values at the time of diagnosis do not appear to correlate with disease severity at onset.

Comparison of discrimination and calibration performance of ECG-based machine learning models for prediction of new-onset atrial fibrillation.

BACKGROUND: Machine learning (ML) methods to build prediction models starting from electrocardiogram (ECG) signals are an emerging research field. The aim of the present study is to investigate the performances of two ML approaches based on ECGs for the prediction of new-onset atrial fibrillation (AF), in terms of discrimination, calibration and sample size dependence. METHODS: We trained two models to predict new-onset AF: a convolutional neural network (CNN), that takes as input the raw ECG signals, and an eXtreme Gradient Boosting model (XGB), that uses the signal's extracted features. A penalized logistic regression model (LR) was used as a benchmark. Discrimination was evaluated with the area under the ROC curve, while calibration with the integrated calibration index. We investigated the dependence of models' performances on the sample size and on class imbalance corrections introduced with random under-sampling. RESULTS: CNN's discrimination was the most affected by the sample size, outperforming XGB and LR only around n = 10.000 observations. Calibration showed only a small dependence on the sample size for all the models considered. Balancing the training set with random undersampling did not improve discrimination in any of the models. Instead, the main effect of imbalance corrections was to worsen the models' calibration (for CNN, integrated calibration index from 0.014 [0.01, 0.018] to 0.17 [0.16, 0.19]). The sample size emerged as a fundamental point for developing the CNN model, especially in terms of discrimination (AUC = 0.75 [0.73, 0.77] when n = 10.000, AUC = 0.80 [0.79, 0.81] when n = 150.000). The effect of the sample size on the other two models was weaker. Imbalance corrections led to poorly calibrated models, for all the approaches considered, reducing the clinical utility of the models. CONCLUSIONS: Our results suggest that the choice of approach in the analysis of ECG should be based on the amount of data available, preferring more standard models for small datasets. Moreover, imbalance correction methods should be avoided when developing clinical prediction models, where calibration is crucial.

Nintedanib in Idiopathic Pulmonary Fibrosis: Tolerability and Safety in a Real Life Experience in a Single Centre in Patients also Treated with Oral Anticoagulant Therapy.

Idiopathic pulmonary fibrosis (IPF) is a rare and severe disease with a median survival of ~3 years. Nintedanib (NTD) has been shown to be useful in controlling interstitial lung disease (ILD) in IPF. Here we describe the experience of NTD use in IPF in a real-life setting. Objective. Our objective was to examine the safety profile and efficacy of nintedanib even in subjects treated with anticoagulants. Clinical data of patients with IPF treated with NTD at our center were retrospectively evaluated at baseline and at 6 and 12 months after the introduction of NTD. The following parameters were recorded: IPF clinical features, NTD tolerability, and pulmonary function tests (PFT) (i.e., Forced Vital Capacity (FVC) and carbon monoxide diffusing capacity (DLCO)). In total, 56 IPF patients (34% female and 66% male, mean onset age: 71 ± 11 years, mean age at baseline: 74 ± 9 years) treated with NTD were identified. At enrollment, HRCT showed an UIP pattern in 45 (80%) and a NSIP in 11 (20%) patients. For FVC and FEV1 we found no significant change between baseline and 6 months, but for DLCO we observed a decrease (p = 0.012). We identified a significant variation between baseline and 12 months for FEV1 (p = 0.039) and for DLCO (p = 0.018). No significant variation was observed for FVC. In the cohort, 18 (32%) individuals suspended NTD and 10 (18%) reduced the dosage. Among individuals that suspended the dosage, 14 (78%) had gastrointestinal (GI) collateral effects (i.e., diarrhea being the most common complaint (67%), followed by nausea/vomiting (17%) and weight loss (6%). Bleeding episodes have also not been reported in patients taking anticoagulant therapy. (61%). One patient died within the first 6 months and two subjects died within the first 12 months. In a real-life clinical scenario, NTD may stabilize the FVC values in IPF patients. However, GI side effects are frequent and NTD dose adjustment may be necessary to retain the drug in IPF patients. This study confirms the safety of NTD, even in patients treated with anticoagulant drugs.

A group of three miRNAs can act as candidate circulating biomarkers in liquid biopsies from melanoma patients.

Background: Staging of melanoma and follow up after melanoma diagnosis aims at predicting risk and detecting progression or recurrence at early stage, respectively in order to timely start and/or change treatment. Tumor thickness according to Breslow, status of the sentinel node and value of the lactate dehydrogenase (LDH) are well-established prognostic markers for metastatic risk, but reliable biomarkers identifying early recurrence or candidates who may benefit best from medical treatment are still warranted. Liquid biopsy has emerged to be a suitable method for identifying biomarkers for early cancer diagnosis, prognosis, therapeutic response prediction, and patient follow-up. Liquid biopsy is a blood-based non-invasive procedure that allows analyzing circulating analytes, including extracellular vesicles. Methods: In this study we have explored the use of 7 miRNAs, namely hsa-miR-149-3p, hsa-miR-150-5p, hsa-miR-21-5p, hsa-miR-200c-3p, hsa-miR-134-5p, hsa-miR-144-3p and hsa-miR-221-3p in plasma exosomes to discriminate melanoma patients from controls without melanoma in a cohort of 92 individuals. Results and discussion: Our results showed that three out seven miRNAs, namely hsa-miR-200c-3p, hsa-miR-144-3p and hsa-miR-221-3p were differentially expressed in plasma-derived exosomes from melanoma patients and controls. Furthermore, the expression of the three miRNAs may be a promising ancillary tool as a melanoma biomarker, even for discriminating between nevi and melanoma.

Deep-learning-based prognostic modeling for incident heart failure in patients with diabetes using electronic health records: A retrospective cohort study.

Patients with type 2 diabetes mellitus (T2DM) have more than twice the risk of developing heart failure (HF) compared to patients without diabetes. The present study is aimed to build an artificial intelligence (AI) prognostic model that takes in account a large and heterogeneous set of clinical factors and investigates the risk of developing HF in diabetic patients. We carried out an electronic health records- (EHR-) based retrospective cohort study that included patients with cardiological clinical evaluation and no previous diagnosis of HF. Information consists of features extracted from clinical and administrative data obtained as part of routine medical care. The primary endpoint was diagnosis of HF (during out-of-hospital clinical examination or hospitalization). We developed two prognostic models using (1) elastic net regularization for Cox proportional hazard model (COX) and (2) a deep neural network survival method (PHNN), in which a neural network was used to represent a non-linear hazard function and explainability strategies are applied to estimate the influence of predictors on the risk function. Over a median follow-up of 65 months, 17.3% of the 10,614 patients developed HF. The PHNN model outperformed COX both in terms of discrimination (c-index 0.768 vs 0.734) and calibration (2-year integrated calibration index 0.008 vs 0.018). The AI approach led to the identification of 20 predictors of different domains (age, body mass index, echocardiographic and electrocardiographic features, laboratory measurements, comorbidities, therapies) whose relationship with the predicted risk correspond to known trends in the clinical practice. Our results suggest that prognostic models for HF in diabetic patients may improve using EHRs in combination with AI techniques for survival analysis, which provide high flexibility and better performance with respect to standard approaches.

microRNAs combined to radiomic features as a predictor of complete clinical response after neoadjuvant radio-chemotherapy for locally advanced rectal cancer: a preliminary study.

OBJECTIVE: To define a predictive Artificial Intelligence (AI) algorithm based on the integration of a set of biopsy-based microRNAs expression data and radiomic features to understand their potential impact in predicting clinical response (CR) to neoadjuvant radio-chemotherapy (nRCT). The identification of patients who would truly benefit from nRCT for Locally Advanced Rectal Cancer (LARC) could be crucial for an improvement in a tailored therapy. METHODS: Forty patients with LARC were retrospectively analyzed. An MRI of the pelvis before and after nRCT was performed. In the diagnostic biopsy, the expression levels of 7 miRNAs were measured and correlated with the tumor response rate (TRG), assessed on the surgical sample. The accuracy of complete CR (cCR) prediction was compared for i) clinical predictors; ii) radiomic features; iii) miRNAs levels; and iv) combination of radiomics and miRNAs. RESULTS: Clinical predictors showed the lowest accuracy. The best performing model was based on the integration of radiomic features with miR-145 expression level (AUC-ROC = 0.90). AI algorithm, based on radiomics features and the overexpression of miR-145, showed an association with the TRG class and demonstrated a significant impact on the outcome. CONCLUSION: The pre-treatment identification of responders/NON-responders to nRCT could address patients to a personalized strategy, such as total neoadjuvant therapy (TNT) for responders and upfront surgery for non-responders. The combination of radiomic features and miRNAs expression data from images and biopsy obtained through standard of care has the potential to accelerate the discovery of a noninvasive multimodal approach to predict the cCR after nRCT for LARC.

Prognostic Prediction of Genotype vs Phenotype in Genetic Cardiomyopathies.

BACKGROUND: Diverse genetic backgrounds often lead to phenotypic heterogeneity in cardiomyopathies (CMPs). Previous genotype-phenotype studies have primarily focused on the analysis of a single phenotype, and the diagnostic and prognostic features of the CMP genotype across different phenotypic expressions remain poorly understood. OBJECTIVES: We sought to define differences in outcome prediction when stratifying patients based on phenotype at presentation compared with genotype in a large cohort of patients with CMPs and positive genetic testing. METHODS: Dilated cardiomyopathy (DCM), arrhythmogenic right ventricular cardiomyopathy, left-dominant arrhythmogenic cardiomyopathy, and biventricular arrhythmogenic cardiomyopathy were examined in this study. A total of 281 patients (80% DCM) with pathogenic or likely pathogenic variants were included. The primary and secondary outcomes were: 1) all-cause mortality (D)/heart transplant (HT); 2) sudden cardiac death/major ventricular arrhythmias (SCD/MVA); and 3) heart failure-related death (DHF)/HT/left ventricular assist device implantation (LVAD). RESULTS: Survival analysis revealed that SCD/MVA events occurred more frequently in patients without a DCM phenotype and in carriers of DSP, PKP2, LMNA, and FLNC variants. However, after adjustment for age and sex, genotype-based classification, but not phenotype-based classification, was predictive of SCD/MVA. LMNA showed the worst trends in terms of D/HT and DHF/HT/LVAD. CONCLUSIONS: Genotypes were associated with significant phenotypic heterogeneity in genetic cardiomyopathies. Nevertheless, in our study, genotypic-based classification showed higher precision in predicting the outcome of patients with CMP than phenotype-based classification. These findings add to our current understanding of inherited CMPs and contribute to the risk stratification of patients with positive genetic testing.

Correlation between Microvascular Damage and Internal Organ Involvement in Scleroderma: Focus on Lung Damage and Endothelial Dysfunction.

BACKGROUND: Systemic sclerosis (SSc) is an incurable connective tissue disease characterized by decreased peripheral blood perfusion due to microvascular damage and skin thickening/hardening. The microcirculation deficit is typically secondary to structural vessel damage, which can be assessed morphologically and functionally in a variety of ways, exploiting different technologies. OBJECTIVE: This paper focuses on reviewing new studies regarding the correlation between microvascular damage, endothelial dysfunction, and internal organ involvement, particularly pulmonary changes in SSc. METHODS: We critically reviewed the most recent literature on the correlation between blood perfusion and organ involvement. RESULTS: Many papers have demonstrated the link between structural microcirculatory damage and pulmonary involvement; however, studies that have investigated correlations between microvascular functional impairment and internal organ damage are scarce. Overall, the literature supports the correlation between organ involvement and functional microcirculatory impairment in SSc patients. CONCLUSIONS: Morphological and functional techniques appear to be emerging biomarkers in SSc, but obviously need further investigation.

Interpretability of time-series deep learning models: A study in cardiovascular patients admitted to Intensive care unit.

Interpretability is fundamental in healthcare problems and the lack of it in deep learning models is currently the major barrier in the usage of such powerful algorithms in the field. The study describes the implementation of an attention layer for Long Short-Term Memory (LSTM) neural network that provides a useful picture on the influence of the several input variables included in the model. A cohort of 10,616 patients with cardiovascular diseases is selected from the MIMIC III dataset, an openly available database of electronic health records (EHRs) including all patients admitted to an ICU at Boston's Medical Centre. For each patient, we consider a 10-length sequence of 1-hour windows in which 48 clinical parameters are extracted to predict the occurrence of death in the next 7 days. Inspired from the recent developments in the field of attention mechanisms for sequential data, we implement a recurrent neural network with LSTM cells incorporating an attention mechanism to identify features driving model's decisions over time. The performance of the LSTM model, measured in terms of AUC, is 0.790 (SD = 0.015). Regard our primary objective, i.e. model interpretability, we investigate the role of attention weights. We find good correspondence with driving predictors of a transparent model (r = 0.611, 95% CI [0.395, 0.763]). Moreover, most influential features identified at the cohort-level emerge as known risk factors in the clinical context. Despite the limitations of study dataset, this work brings further evidence of the potential of attention mechanisms in making deep learning model more interpretable and suggests the application of this strategy for the sequential analysis of EHRs.

DKC1 Overexpression Induces a More Aggressive Cellular Behavior and Increases Intrinsic Ribosomal Activity in Immortalized Mammary Gland Cells.

Dyskerin is a nucleolar protein involved in the small nucleolar RNA (snoRNA)-guided pseudouridylation of specific uridines on ribosomal RNA (rRNA), and in the stabilization of the telomerase RNA component (hTR). Loss of function mutations in DKC1 causes X-linked dyskeratosis congenita, which is characterized by a failure of proliferating tissues and increased susceptibility to cancer. However, several tumors show dyskerin overexpression. We observed that patients with primary breast cancers with high dyskerin levels are more frequently characterized by shorter survival rates and positive lymph node status than those with tumors with a lower dyskerin expression. To functionally characterize the effects of high dyskerin expression, we generated stably overexpressing DKC1 models finding that increased dyskerin levels conferred a more aggressive cellular phenotype in untransformed immortalized MCF10A cells. Contextually, DKC1 overexpression led to an upregulation of some snoRNAs, including SNORA67 and a significantly increased U1445 modification on 18S rRNA, the known target of SNORA67. Lastly, we found that dyskerin overexpression strongly enhanced the synthetic activity of ribosomes increasing translational efficiency in MCF10A. Altogether, our results indicate that dyskerin may sustain the neoplastic phenotype from an early stage in breast cancer endowing ribosomes with an augmented translation efficiency.

A bird's-eye view of Italian genomic variation through whole-genome sequencing.

The genomic variation of the Italian peninsula populations is currently under characterised: the only Italian whole-genome reference is represented by the Tuscans from the 1000 Genome Project. To address this issue, we sequenced a total of 947 Italian samples from three different geographical areas. First, we defined a new Italian Genome Reference Panel (IGRP1.0) for imputation, which improved imputation accuracy, especially for rare variants, and we tested it by GWAS analysis on red blood traits. Furthermore, we extended the catalogue of genetic variation investigating the level of population structure, the pattern of natural selection, the distribution of deleterious variants and occurrence of human knockouts (HKOs). Overall the results demonstrate a high level of genomic differentiation between cohorts, different signatures of natural selection and a distinctive distribution of deleterious variants and HKOs, confirming the necessity of distinct genome references for the Italian population.

Associations of autozygosity with a broad range of human phenotypes.

In many species, the offspring of related parents suffer reduced reproductive success, a phenomenon known as inbreeding depression. In humans, the importance of this effect has remained unclear, partly because reproduction between close relatives is both rare and frequently associated with confounding social factors. Here, using genomic inbreeding coefficients (FROH) for >1.4 million individuals, we show that FROH is significantly associated (p < 0.0005) with apparently deleterious changes in 32 out of 100 traits analysed. These changes are associated with runs of homozygosity (ROH), but not with common variant homozygosity, suggesting that genetic variants associated with inbreeding depression are predominantly rare. The effect on fertility is striking: FROH equivalent to the offspring of first cousins is associated with a 55% decrease [95% CI 44-66%] in the odds of having children. Finally, the effects of FROH are confirmed within full-sibling pairs, where the variation in FROH is independent of all environmental confounding.

Can we predict firms' innovativeness? The identification of innovation performers in an Italian region through a supervised learning approach.

The study shows the feasibility of predicting firms' expenditures in innovation, as reported in the Community Innovation Survey, applying a supervised machine-learning approach on a sample of Italian firms. Using an integrated dataset of administrative records and balance sheet data, designed to include all informative variables related to innovation but also easily accessible for most of the cohort, random forest algorithm is implemented to obtain a classification model aimed to identify firms that are potential innovation performers. The performance of the classifier, estimated in terms of AUC, is 0.794. Although innovation investments do not always result in patenting, the model is able to identify 71.92% of firms with patents. More encouraging results emerge from the analysis of the inner working of the model: predictors identified as most important-such as firm size, sector belonging and investment in intangible assets-confirm previous findings of literature, but in a completely different framework. The outcomes of this study are considered relevant for both economic analysts, because it demonstrates the potential of data-driven models for understanding the nature of innovation behaviour, and practitioners, such as policymakers or venture capitalists, who can benefit by evidence-based tools in the decision-making process.

Publisher Correction: Genome-wide association meta-analysis of individuals of European ancestry identifies new loci explaining a substantial fraction of hair color variation and heritability.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

A multi-ancestry genome-wide study incorporating gene-smoking interactions identifies multiple new loci for pulse pressure and mean arterial pressure.

Elevated blood pressure (BP), a leading cause of global morbidity and mortality, is influenced by both genetic and lifestyle factors. Cigarette smoking is one such lifestyle factor. Across five ancestries, we performed a genome-wide gene-smoking interaction study of mean arterial pressure (MAP) and pulse pressure (PP) in 129 913 individuals in stage 1 and follow-up analysis in 480 178 additional individuals in stage 2. We report here 136 loci significantly associated with MAP and/or PP. Of these, 61 were previously published through main-effect analysis of BP traits, 37 were recently reported by us for systolic BP and/or diastolic BP through gene-smoking interaction analysis and 38 were newly identified (P < 5 × 10-8, false discovery rate < 0.05). We also identified nine new signals near known loci. Of the 136 loci, 8 showed significant interaction with smoking status. They include CSMD1 previously reported for insulin resistance and BP in the spontaneously hypertensive rats. Many of the 38 new loci show biologic plausibility for a role in BP regulation. SLC26A7 encodes a chloride/bicarbonate exchanger expressed in the renal outer medullary collecting duct. AVPR1A is widely expressed, including in vascular smooth muscle cells, kidney, myocardium and brain. FHAD1 is a long non-coding RNA overexpressed in heart failure. TMEM51 was associated with contractile function in cardiomyocytes. CASP9 plays a central role in cardiomyocyte apoptosis. Identified only in African ancestry were 30 novel loci. Our findings highlight the value of multi-ancestry investigations, particularly in studies of interaction with lifestyle factors, where genomic and lifestyle differences may contribute to novel findings.

Multiancestry Genome-Wide Association Study of Lipid Levels Incorporating Gene-Alcohol Interactions.

A person's lipid profile is influenced by genetic variants and alcohol consumption, but the contribution of interactions between these exposures has not been studied. We therefore incorporated gene-alcohol interactions into a multiancestry genome-wide association study of levels of high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, and triglycerides. We included 45 studies in stage 1 (genome-wide discovery) and 66 studies in stage 2 (focused follow-up), for a total of 394,584 individuals from 5 ancestry groups. Analyses covered the period July 2014-November 2017. Genetic main effects and interaction effects were jointly assessed by means of a 2-degrees-of-freedom (df) test, and a 1-df test was used to assess the interaction effects alone. Variants at 495 loci were at least suggestively associated (P < 1 × 10-6) with lipid levels in stage 1 and were evaluated in stage 2, followed by combined analyses of stage 1 and stage 2. In the combined analysis of stages 1 and 2, a total of 147 independent loci were associated with lipid levels at P < 5 × 10-8 using 2-df tests, of which 18 were novel. No genome-wide-significant associations were found testing the interaction effect alone. The novel loci included several genes (proprotein convertase subtilisin/kexin type 5 (PCSK5), vascular endothelial growth factor B (VEGFB), and apolipoprotein B mRNA editing enzyme, catalytic polypeptide 1 (APOBEC1) complementation factor (A1CF)) that have a putative role in lipid metabolism on the basis of existing evidence from cellular and experimental models.

Genome-wide association meta-analysis of 30,000 samples identifies seven novel loci for quantitative ECG traits.

Genome-wide association studies (GWAS) of quantitative electrocardiographic (ECG) traits in large consortia have identified more than 130 loci associated with QT interval, QRS duration, PR interval, and heart rate (RR interval). In the current study, we meta-analyzed genome-wide association results from 30,000 mostly Dutch samples on four ECG traits: PR interval, QRS duration, QT interval, and RR interval. SNP genotype data was imputed using the Genome of the Netherlands reference panel encompassing 19 million SNPs, including millions of rare SNPs (minor allele frequency < 5%). In addition to many known loci, we identified seven novel locus-trait associations: KCND3, NR3C1, and PLN for PR interval, KCNE1, SGIP1, and NFKB1 for QT interval, and ATP2A2 for QRS duration, of which six were successfully replicated. At these seven loci, we performed conditional analyses and annotated significant SNPs (in exons and regulatory regions), demonstrating involvement of cardiac-related pathways and regulation of nearby genes.

Genome-wide association meta-analysis of individuals of European ancestry identifies new loci explaining a substantial fraction of hair color variation and heritability.

Hair color is one of the most recognizable visual traits in European populations and is under strong genetic control. Here we report the results of a genome-wide association study meta-analysis of almost 300,000 participants of European descent. We identified 123 autosomal and one X-chromosome loci significantly associated with hair color; all but 13 are novel. Collectively, single-nucleotide polymorphisms associated with hair color within these loci explain 34.6% of red hair, 24.8% of blond hair, and 26.1% of black hair heritability in the study populations. These results confirm the polygenic nature of complex phenotypes and improve our understanding of melanin pigment metabolism in humans.

A Large-Scale Multi-ancestry Genome-wide Study Accounting for Smoking Behavior Identifies Multiple Significant Loci for Blood Pressure.

Genome-wide association analysis advanced understanding of blood pressure (BP), a major risk factor for vascular conditions such as coronary heart disease and stroke. Accounting for smoking behavior may help identify BP loci and extend our knowledge of its genetic architecture. We performed genome-wide association meta-analyses of systolic and diastolic BP incorporating gene-smoking interactions in 610,091 individuals. Stage 1 analysis examined ∼18.8 million SNPs and small insertion/deletion variants in 129,913 individuals from four ancestries (European, African, Asian, and Hispanic) with follow-up analysis of promising variants in 480,178 additional individuals from five ancestries. We identified 15 loci that were genome-wide significant (p < 5 × 10-8) in stage 1 and formally replicated in stage 2. A combined stage 1 and 2 meta-analysis identified 66 additional genome-wide significant loci (13, 35, and 18 loci in European, African, and trans-ancestry, respectively). A total of 56 known BP loci were also identified by our results (p < 5 × 10-8). Of the newly identified loci, ten showed significant interaction with smoking status, but none of them were replicated in stage 2. Several loci were identified in African ancestry, highlighting the importance of genetic studies in diverse populations. The identified loci show strong evidence for regulatory features and support shared pathophysiology with cardiometabolic and addiction traits. They also highlight a role in BP regulation for biological candidates such as modulators of vascular structure and function (CDKN1B, BCAR1-CFDP1, PXDN, EEA1), ciliopathies (SDCCAG8, RPGRIP1L), telomere maintenance (TNKS, PINX1, AKTIP), and central dopaminergic signaling (MSRA, EBF2).