Citations

Purpose This study aimed to compare accuracy and efficiency of a convolutional neural network (CNN)–enhanced workflow for pancreas segmentation versus radiologists in the context of interreader reliability. Methods Volumetric pancreas segmentations on a data set of 294 portal venous computed tomographies were performed by 3 radiologists (R1, R2, and R3) and by a CNN. Convolutional neural network segmentations were reviewed and, if needed, corrected (“corrected CNN [c-CNN]” segmentations) by radiologists. Ground truth was obtained from radiologists' manual segmentations using simultaneous truth and performance level estimation algorithm. Interreader reliability and model's accuracy were evaluated with Dice-Sorenson coefficient (DSC) and Jaccard coefficient (JC). Equivalence was determined using a two 1-sided test. Convolutional neural network segmentations below the 25th percentile DSC were reviewed to evaluate segmentation errors. Time for manual segmentation and c-CNN was compared. Results Pancreas volumes from 3 sets of segmentations (manual, CNN, and c-CNN) were noninferior to simultaneous truth and performance level estimation–derived volumes [76.6 cm3 (20.2 cm3), P < 0.05]. Interreader reliability was high (mean [SD] DSC between R2-R1, 0.87 [0.04]; R3-R1, 0.90 [0.05]; R2-R3, 0.87 [0.04]). Convolutional neural network segmentations were highly accurate (DSC, 0.88 [0.05]; JC, 0.79 [0.07]) and required minimal-to-no corrections (c-CNN: DSC, 0.89 [0.04]; JC, 0.81 [0.06]; equivalence, P < 0.05). Undersegmentation (n = 47 [64%]) was common in the 73 CNN segmentations below 25th percentile DSC, but there were no major errors. Total inference time (minutes) for CNN was 1.2 (0.3). Average time (minutes) taken by radiologists for c-CNN (0.6 [0.97]) was substantially lower compared with manual segmentation (3.37 [1.47]; savings of 77.9%–87% [P < 0.0001]). Conclusions Convolutional neural network–enhanced workflow provides high accuracy and efficiency for volumetric pancreas segmentation on computed tomography.

BackgroundCurative surgical treatments afford the best prognosis for patients with intrahepatic cholangiocarcinoma (iCCA); however, the comparative effectiveness of treatment options and factors associated with curative treatment receipt for early stage iCCA remain unknown.MethodsThe authors identified patients who were diagnosed with early stage iCCA, defined as a unifocal tumor <3 cm, during 2004–2018 from the National Cancer Database. Multivariable logistic and Cox regression analyses were used to identify the factors associated with curative treatment and overall survival (OS), respectively.ResultsThe proportion of patients with early stage iCCA increased from 4.5% in 2004 to 7.3% in 2018, with the odds of early stage detection increasing by 3.1% per year (odds ratio [OR], 1.031; 95% CI, 1.015–1.049). Of 1093 patients who had early stage iCCA, 464 (42.5%) underwent resection, 113 (10.3%) underwent ablation, 62 (5.7%) underwent liver transplantation, and 454 (41.5%) received noncurative treatments. Hispanic patients (adjusted OR [aOR], 0.57; 95% CI, 0.33–0.97) and Black patients (aOR, 0.47; 95% CI, 0.28–0.77) were less likely to receive curative treatments than White patients. Compared with patients who underwent surgical resection, those who underwent liver transplantation had a trend toward improved OS (adjusted hazard ratio [aHR], 0.63; 95% CI, 0.37–1.08), whereas those who underwent local ablation (aHR, 1.39; 95% CI, 1.01–1.92) and noncurative treatments (aHR, 3.97; 95% CI, 3.24–4.88) experienced worse OS.ConclusionsMore than one third of patients with early stage iCCA did not receive curative treatment, with Hispanic and Black patients being less likely to receive curative treatments than White patients. Surgical resection and liver transplantation were associated with improved survival compared with local ablation. Future studies should investigate disparities in curative treatment receipt and outcomes for early stage iCCA.

Liver transplantation (LT) is a life‐saving treatment for patients with acute liver failure (ALF). Currently, there are few detailed data regarding long‐term outcomes after LT for ALF. We combined prospective data from the Acute Liver Failure Study Group (ALFSG) Registry with those of the Scientific Registry of Transplant Recipients (SRTR) to assess outcomes among consecutive patients with ALF listed for LT. Cohort analysis of detailed pretransplantation data for patients listed for LT for ALF in the ALFSG Registry between January 1998 and October 2018 matched with transplantation‐related data from the SRTR. Primary outcomes were 1‐ and 3‐year post‐LT patient survival. Secondary outcome was receipt of LT; independent associations with successful receipt of LT were determined using multivariable logistic regression. Of 624 patients with ALF listed for LT, 398 (64%) underwent LT, 100 (16%) died without LT, and 126 (20%) recovered spontaneously. Among LT recipients, etiologies included seronegative/indeterminate (22%), drug‐induced liver injury (18%), acetaminophen overdose (APAP; 16%), and viral hepatitis (15%). The 1‐ and 3‐year post‐LT patient survival rates were 91% and 90%, respectively. Comparing those dying on the waiting list versus with those who received LT, the former had more severe multiorgan failure, reflected by increased vasopressor use (65% vs. 22%), mechanical ventilation (84% vs. 57%), and renal replacement therapy (57% vs. 30%; p < 0.0001 for all). After adjusting for relevant covariates, age (adjusted odds ratio [aOR] 1.02, 95% confidence interval [CI] 1.00–1.04), APAP etiology (aOR 2.72, 95% CI 1.42–5.23), requirement for vasopressors (aOR 4.19, 95% CI 2.44–7.20), Grade III/IV hepatic encephalopathy (aOR 2.47, 95% CI 1.29–4.72), and Model for End‐Stage Liver Disease (MELD) scores (aOR 1.05, 95% CI 1.02–1.09; p < 0.05 for all) were independently associated with death without receipt of LT. Post‐LT outcomes for ALF are excellent in this cohort of very ill patients. The development of multiorgan failure while on the transplantation list and APAP ALF etiology were associated with a lower likelihood of successful receipt of LT.

AbstractBackgroundTransarterial radioembolization (TARE) is increasingly used as an alternative to transarterial chemoembolization (TACE) for the treatment of hepatocellular carcinoma (HCC). We aimed to perform an overall and individual patient data (IPD) meta‐analysis of studies comparing TACE and TARE.MethodsWe performed a systematic literature search using pre‐specified keywords with the aid of an informationist for articles from inception to 3/2020. The primary endpoint was overall survival (OS), and the secondary endpoint was time to progression (TTP).ResultsSeventeen studies met inclusion criteria with 2465 unique patients, with one randomized trial, 4 prospective studies and 12 retrospective studies. Barcelona Clinic Liver Cancer (BCLC) stage B (42.8%) was the most common stage followed by BCLC A (30.3%) and BCLC C (29.0%). There was no difference in OS between the two modalities (−0.55 months, 95% CI −1.95 to 3.05). In three studies with available TTP data, TARE resulted in a longer TTP than TACE (mean TTP 17.5 vs. 9.8 months; mean TTP difference 4.8 months, 95% CI 1.3–8.3 months). IPD‐level meta‐analysis of 311 patients from three studies showed no difference in overall OS between the two modalities including among subgroups stratified by tumor stage and liver function. Limitations of the current literature include inconsistent length of follow‐up, inconsistency in response criteria, and safety reporting.ConclusionsCurrent data suggest TARE provides significantly longer TTP than TACE, although the two treatments do not significantly differ in terms of OS. Given limitations of the current data, there is rationale for prospective studies comparing these modalities.

Abstract Intratumoral heterogeneity arising from tumor evolution poses significant challenges biologically and clinically. Dissecting this complexity may benefit from deep learning (DL) algorithms, which can infer molecular features from ubiquitous hematoxylin and eosin (H&E)–stained tissue sections. Although DL algorithms have been developed to predict some driver mutations from H&E images, the ability of these DL algorithms to resolve intratumoral mutation heterogeneity at subclonal spatial resolution is unexplored. Here, we apply DL to a paradigm of intratumoral heterogeneity, clear cell renal cell carcinoma (ccRCC), the most common type of kidney cancer. Matched IHC and H&E images were leveraged to develop DL models for predicting intratumoral genetic heterogeneity of the three most frequently mutated ccRCC genes, BAP1, PBRM1, and SETD2. DL models were generated on a large cohort (N = 1,282) and tested on several independent cohorts, including a TCGA cohort (N = 363 patients) and two tissue microarray (TMA) cohorts (N = 118 and 365 patients). These models were also expanded to a patient-derived xenograft (PDX) TMA, affording analysis of homotopic and heterotopic interactions of tumor and stroma. The status of all three genes could be inferred by DL, with BAP1 showing the highest sensitivity and performance within and across tissue samples (AUC = 0.87–0.89 on holdout). BAP1 results were validated on independent human (AUC = 0.77–0.84) and PDX (AUC = 0.80) cohorts. Finally, BAP1 predictions correlated with clinical outputs such as disease-specific survival. Overall, these data show that DL models can resolve intratumoral heterogeneity in cancer with potential diagnostic, prognostic, and biological implications. Significance: This work demonstrates the potential for deep learning analysis of histopathologic images to serve as a fast, low-cost method to assess genetic intratumoral heterogeneity. See related commentary by Song et al., p. 2672

Abstract Background Nearly half of patients with hidradenitis suppurativa (HS) report dissatisfaction with their treatment. However, factors related to treatment satisfaction have not been explored. Objectives To measure associations between treatment satisfaction and clinical and treatment-related characteristics among patients with HS. Methods Treatment satisfaction was evaluated utilizing data from a cross-sectional global survey of patients with HS recruited from 27 institutions, mainly HS referral centres, in 14 different countries from October 2017 to July 2018. The primary outcome was patients’ self-reported overall satisfaction with their current treatments for HS, rated on a five-point scale from ‘very dissatisfied’ to ‘very satisfied’. Results The final analysis cohort comprised 1418 patients with HS, most of whom were European (55%, 780 of 1418) or North American (38%, 542 of 1418), and female (85%, 1210 of 1418). Overall, 45% (640 of 1418) of participants were either dissatisfied or very dissatisfied with their current medical treatment. In adjusted analysis, patients primarily treated by a dermatologist for HS had 1·99 [95% confidence interval (CI) 1·62–2·44, P &lt; 0·001] times the odds of being satisfied with current treatment than participants not primarily treated by a dermatologist. Treatment with biologics was associated with higher satisfaction [odds ratio (OR) 2·36, 95% CI 1·74–3·19, P &lt; 0·001] relative to treatment with nonbiologic systemic medications. Factors associated with lower treatment satisfaction included smoking (OR 0·78, 95% CI 0·62–0·99; active vs. never), depression (OR 0·69, 95% CI 0·54–0·87), increasing number of comorbidities (OR 0·88 per comorbidity, 95% CI 0·81–0·96) and increasing flare frequency. Conclusions There are several factors that appear to positively influence satisfaction with treatment among patients with HS, including treatment by a dermatologist and treatment with a biologic medication. Factors that appear to lower treatment satisfaction include active smoking, depression, accumulation of comorbid conditions and increasing flare frequency. Awareness of these factors may support partnered decision making with the goal of improving treatment outcomes. What is already known about this topic? Nearly half of patients with hidradenitis suppurativa report dissatisfaction with their treatments. What does this study add? Satisfaction with treatment is increased by receiving care from a dermatologist and treatment with biologics.Satisfaction with treatment is decreased by tobacco smoking, accumulation of comorbid conditions including depression, and higher flare frequency. What are the clinical implications of this work? Awareness of the identified factors associated with poor treatment satisfaction may support partnered decision making and improve treatment outcomes.

Abstract Cirrhosis is complicated by a high rate of nosocomial infections (NIs), which result in poor outcomes and are challenging to predict using clinical variables alone. Our aim was to determine predictors of NI using admission serum metabolomics and gut microbiota in inpatients with cirrhosis. In this multicenter inpatient cirrhosis study, serum was collected on admission for liquid chromatography–mass spectrometry metabolomics, and a subset provided stool for 16SrRNA analysis. Hospital course, including NI development and death, were analyzed. Metabolomic analysis using analysis of covariance (ANCOVA) (demographics, Model for End‐Stage Liver Disease [MELD] admission score, white blood count [WBC], rifaximin, and infection status adjusted) and random forest analyses for NI development were performed. Additional values of serum metabolites over clinical variables toward NI were evaluated using logistic regression. Stool microbiota and metabolomic correlations were compared in patients with and without NI development. A total of 602 patients (231 infection admissions) were included; 101 (17%) developed NIs, which resulted in worse inpatient outcomes, including intensive care unit transfer, organ failure, and death. A total of 127 patients also gave stool samples, and 20 of these patients developed NIs. The most common NIs were spontaneous bacterial peritonitis followed by urinary tract infection, Clostridioides difficile, and pneumonia. A total of 247 metabolites were significantly altered on ANCOVA. Higher MELD scores (odds ratio, 1.05; p < 0.0001), admission infection (odds ratio, 3.54; p < 0.0001), and admission WBC (odds ratio, 1.05; p = 0.04) predicted NI (area under the curve, 0.74), which increased to 0.77 (p = 0.05) with lower 1‐linolenoyl‐glycerolphosphocholine (GPC) and 1‐stearoyl‐GPC and higher N‐acetyltryptophan and N‐acetyl isoputreanine. Commensal microbiota were lower and pathobionts were higher in those who developed NIs. Microbial–metabolite correlation networks were complex and dense in patients with NIs, especially sub‐networks centered on Ruminococcaceae and Pseudomonadaceae. NIs are common and associated with poor outcomes in cirrhosis. Admission gut microbiota in patients with NIs showed higher pathobionts and lower commensal microbiota. Microbial–metabolomic correlations were more complex, dense, and homogeneous among those who developed NIs, indicating greater linkage strength. Serum metabolites and gut microbiota on admission are associated with NI development in cirrhosis.

Introduction The Accreditation Council for Graduate Medical Education (ACGME) offers descriptions of competencies and milestones but does not provide standardized assessments to track trainee competency. Entrustable professional activities (EPAs) and special assessments (SAs) are emerging methods to assess the level of competency obtained by regional anesthesiology and acute pain medicine (RAAPM) fellows. Methods A panel of RAAPM physicians with experience in education and competency assessment and one medical student were recruited to participate in a modified Delphi method with iterative rounds to reach consensus on: a list of EPAs, SAs, and procedural skills; detailed definitions for each EPA and SA; a mapping of the EPAs and SAs to the ACGME milestones; and a target level of entrustment for graduating US RAAPM fellows for each EPA and procedural skill. A gap analysis was performed and a heat map was created to cross-check the EPAs and SAs to the ACGME milestones. Results Participants in EPA and SA development included 19 physicians and 1 medical student from 18 different programs. The Delphi rounds yielded a final list of 23 EPAs, a defined entrustment scale, mapping of the EPAs to ACGME milestones, and graduation targets. A list of 73 procedural skills and 7 SAs were similarly developed. Discussion A list of 23 RAAPM EPAs, 73 procedural skills, and 7 SAs were created using a rigorous methodology to reach consensus. This framework can be utilized to help assess RAAPM fellows in the USA for competency and allow for meaningful performance feedback.