Citations

AimsMachine learning in digital pathology can improve efficiency and accuracy via prescreening with automated feature identification. Studies using uniform histological material have shown promise. Generalised application requires validation on slides from multiple institutions. We used machine learning to identify glomeruli on renal biopsies and compared performance between single and multiple institutions.Methods and resultsRandomly selected, adequately sampled renal core biopsy cases (71) consisting of four stains each (haematoxylin and eosin, trichrome, silver, periodic acid Schiff) from three institutions were digitised at ×40. Glomeruli were manually annotated by three renal pathologists using a digital tool. Cases were divided into training/validation (n = 52) and evaluation (n = 19) cohorts. An algorithm was trained to develop three convolutional neural network (CNN) models which tested case cohorts intra‐ and inter‐institutionally. Raw CNN search data from each of the four slides per case were merged into composite regions of interest containing putative glomeruli. The sensitivity and modified specificity of glomerulus detection (versus annotated truth) were calculated for each model/cohort. Intra‐institutional (3) sensitivity ranged from 90 to 93%, with modified specificity from 86 to 98%. Interinstitutional (1) sensitivity was 77%, with modified specificity 97%. Combined intra‐ and inter‐institutional (1) sensitivity was 86%, with modified specificity 92%.ConclusionsFeature detection sensitivity degrades when training and test material originate from different sites. Training using a combined set of digital slides from three institutions improves performance. Differing histology methods probably account for algorithm performance contrasts. Our data highlight the need for diverse training sets for the development of generalisable machine learning histology algorithms.

Abstract Background There is a critical need to identify patient characteristics associated with long-term ovarian cancer survival. Methods Quality of life (QOL), measured by the Functional Assessment of Cancer Therapy-Ovarian-Trial Outcome Index (FACT-O-TOI), including physical, functional, and ovarian-specific subscales, was compared between long-term survivors (LTS) (8+ years) and short-term survivors (STS) (<5 years) of GOG 218 at baseline; before cycles 4, 7, 13, 21; and 6 months post-treatment using linear and longitudinal mixed models adjusted for covariates. Adverse events (AEs) were compared between survivor groups at each assessment using generalized linear models. All P values are 2-sided. Results QOL differed statistically significantly between STS (N = 1115) and LTS (N = 260) (P < .001). Baseline FACT-O-TOI and FACT-O-TOI change were independently associated with long-term survival (odds ratio = 1.05, 95% confidence interval = 1.03 to 1.06 and odds ratio = 1.06, 95% confidence interval = 1.05 to 1.07, respectively). A 7-point increase in baseline QOL was associated with a 38.0% increase in probability of LTS, and a 9-point increase in QOL change was associated with a 67.0% increase in odds for LTS. QOL decreased statistically significantly with increasing AE quartiles (cycle 4 quartiles: 0-5 vs 6-8 vs 9-11 vs ≥12 AEs, P = .01; cycle 21 quartiles: 0-2 vs 3 vs 4-5 vs ≥6 AEs, P = .001). Further, LTS reported statistically significantly better QOL compared with STS (P = .03 and P = .01, cycles 4 and 21, respectively), with similar findings across higher AE grades. Conclusions Baseline and longitudinal QOL change scores distinguished LTS vs STS and are robust prognosticators for long-term survival. Results have trial design and supportive care implications, providing meaningful prognostic value in this understudied population.

AbstractKidney transplant (KT) outcomes from high kidney donor profile index (KDPI ≥85%) donors with acute kidney injury (AKI) remain underreported. KT from 172 high KDPI Acute Kidney Injury Network (AKIN) stage 0‐1 donors and 76 high KDPI AKIN stage 2‐3 donors from a single center were retrospectively assessed. The AKIN 2‐3 cohort had more delayed graft function (71% vs. 37%, p < .001). At one year, there were no differences in the estimated glomerular filtration rate (44 ± 17 vs. 46 ± 18, p = .42) or fibrosis on protocol biopsy (ci, p = .85). Donor terminal creatinine (p = .59) and length of delayed graft function (p = .39) did not impact one‐year eGFR. There were more primary nonfunction (PNF) events in the high KDPI AKIN 2‐3 group (5.3% vs. 0.6%, p = .02). With a median follow‐up of 3.8 years, one‐year death‐censored graft failure was 3.5% for AKIN 0‐1 and 14.5% for AKIN 2‐3 (HR 2.40, 95% CI 1.24‐4.63, p = .01). Although AKIN stage 2‐3 high KDPI kidneys had comparable one‐year eGFR to AKIN stage 0‐1 high KDPI kidneys, there were more PNF occurrences and one‐year death‐censored graft survival was reduced. Given these findings, additional precautions should be undertaken when assessing and utilizing kidneys from severe AKI high KDPI donors.

Background: To evaluate differences in the proportion of uterine fibroid (UF) treatments that are uterine-sparing between Black women and White women and identify factors that could explain disparities. Methods: Women at age 18–54 years who were enrolled from 10 clinical sites in the United States into the Comparing Options for Management: Patient-Centered Results for UFs (COMPARE-UF) treatment registry completed questionnaires before their UF procedure. UF symptoms and quality of life were assessed by questionnaires. Details on UF imaging and treatment (hysterectomy, myomectomy, or uterine artery embolization [UAE]) were collected from each patient's medical record. Random-effects logistic regression was used to assess the association between race and the odds of having a uterine-sparing procedure versus hysterectomy. Subgroup analyses compared each uterine-sparing procedure with hysterectomy. Results: In this cohort of 1141 White women and 1196 Black women, Black women tended to be younger (median 41.0 vs. 42.0 years) and report worse symptoms, pain, and function on every scale compared with White women. Black women were more likely to have had a prior UF treatment compared with White women (22.8% vs. 14.6%). White women had more hysterectomies (43.6% vs. 32.2%) and myomectomies (50.9% vs. 50.2%) versus Black women. Black women had more UAEs (15.1% vs. 4.7%) than White women. After adjusting for clinical site and other variables, Black women had greater odds than White women of having a myomectomy (odds ratio [OR] = 2.41, 95% confidence interval [CI] = 1.63–3.56) or a UAE versus hysterectomy (OR = 4.24, 95% CI = 2.41–7.46). Conclusion: In these participants, Black women were more likely to schedule a uterine-sparing UF treatment and a nonsurgical UF treatment than their White counterparts; this may not be true for all women. Longer comparative effectiveness studies are needed to inform women about the durability of UF treatments. Greater understanding of factors influencing treatment selection is needed as are studies that include women without access to tertiary care centers. Clinical Trial Registration: Clinicaltrials.gov, NCT02260752 (enrollment start: November 2015).

ObjectiveTo determine if wideband tympanometry (WBT) can differentiate types of middle ear effusion (MEE): serous, mucoid, and purulent.Study DesignProspective cohort study.SettingTertiary care children’s hospital.MethodsChildren who met American Academy of Otolaryngology–Head and Neck Surgery’s guidelines for ventilation tube insertion had WBT after anesthesia induction but before tympanotomy. MEE was categorized into 1 of 4 comparison groups: serous effusion, mucoid effusion, purulent effusion, or no effusion. WBT measurements were averaged to 16 one‐third octave frequency bands, and comparison of the absorbance patterns for each MEE type was performed through a linear mixed effects model.ResultsA total of 118 children (211 ears) were included: 47 females (39.8%) and 71 males (60.2%). The mean age was 2.73 years (95% CI, 2.25‐3.22); mean weight, 14.35 kg (95% CI, 12.85‐15.85); and mean Z score, 1.13 (95% CI, –0.64 to 2.33). Effusions included 61 mucoid (28.9%), 30 purulent (14.2%), and 14 serous (6.6%), with 106 (50.2%) having no effusion. No significant differences were found for sex, race, age, weight, or Z score among the 4 types of effusion (P <. 05). WBT showed a significant difference in median absorption among the effusion groups (P <. 001), with a medium effect size of 0.35.ConclusionsWBT has potential use to differentiate types of MEE and should be studied further as a tool for investigating how the natural history and management of serous and mucoid effusions may differ.

Objective: To investigate the role of intraoperative estimated blood loss (EBL) on development of clinically relevant postoperative pancreatic fistula (CR-POPF) after pancreatoduodenectomy (PD). Background: Minimizing EBL has been shown to decrease transfusions and provide better perioperative outcomes in PD. EBL is also felt to be influential on CR-POPF development. Methods: This study consists of 5534 PDs from a 17-institution collaborative (2003–2018). EBL was progressively categorized (≤150mL; 151–400mL; 401–1,000 mL; > 1,000 mL). Impact of additive EBL was assessed using 20 3– factor fistula risk score (FRS) scenarios reflective of endogenous CR-POPF risk. Results: CR-POPF developed in 13.6% of patients (N = 753) and median EBL was 400 mL (interquartile range 250-600 mL). CR-POPF and Grade C POPF were associated with elevated EBL (median 350 vs 400 mL, P = 0.002; 372 vs 500 mL, P < 0.001, respectively). Progressive EBL cohorts displayed incremental CR-POPF rates (8.5%, 13.4%, 15.2%, 16.9%; P < 0.001). EBL >400mL was associated with increased CR-POPF occurrence in 13/20 endogenous risk scenarios. Moreover, 8 of 10 scenarios predicated on a soft gland demonstrated increased CR-POPF incidence. Hypothetical projections demonstrate significant reductions in CR-POPF can be obtained with 1–, 2–, and 3-point decreases in FRS points attributed to EBL risk (12.2%, 17.4%, and 20.0%; P < 0.001). This is especially pronounced in high-risk (FRS7–10) patients, who demonstrate up to a 31% reduction (P < 0.001). Surgeons in the lowest-quartile of median EBL demonstrated CR-POPF rates less than half those in the upper-quartile (7.9% vs 18.8%; P < 0.001). Conclusion: EBL independently contributes significant biological risk to CR-POPF. Substantial reductions in CR-POPF occurrence are projected and obtainable by minimizing EBL. Decreased individual surgeon EBL is associated with improvements in CR-POPF.

Abstract BACKGROUND Intravenous (IV) alteplase with mechanical thrombectomy has been found to be superior to alteplase alone in select patients with intracranial large vessel occlusion. Current guidelines discourage the use of antiplatelet agents or heparin for 24 h following alteplase. However, their use is often necessary in certain circumstances during thrombectomy procedures. OBJECTIVE To study the safety and outcomes in patients who received blood thinning medications for thrombectomy after IV Tissue-Type plasminogen activator (tPA). METHODS This is a multicenter retrospective review of the use of antiplatelet agents and/or heparin in patients within 24 h following tPA administration. Patient demographics, comorbidities, bleeding complications, and discharge outcomes were collected. RESULTS A series of 88 patients at 9 centers received antiplatelet medications and/or heparin anticoagulation following IV alteplase for revascularization procedures requiring stenting. The mean National Institutes of Health Stroke Scale (NIHSS) on admission was 14.6. Reasons for use of a stent included internal carotid artery occlusion in 74% of patients. Thrombolysis in cerebral infarction (TICI) 2b-3 revascularization was accomplished in 90% of patients. The rate of symptomatic intracranial hemorrhage (sICH) was 8%; this was not significantly different than the sICH rate for a matched group of patients not receiving antiplatelets or heparin during the same time frame. Functional independence at 90 d (modified Rankin Scale 0-2) was seen in 57.8% of patients. All-cause mortality was 12%. CONCLUSION The use of antiplatelet agents and heparin for stroke interventions following IV alteplase appears to be safe without significant increased risk of hemorrhagic complications in this group of patients when compared to control data and randomized controlled trials.

Background: Allogeneic HCT continues to be the optimal consolidation strategy for many patients with AML. Cytogenetic and molecular abnormalities are known to influence post-transplant outcomes. We tested the prognostic ability of a modified (mELN) classification system based on available CIBMTR genetic data, to predict post-transplant outcomes. Methods: Adult patients with a diagnosis of AML in first complete remission (CR1) with available pre-transplant cytogenetic and molecular mutational data, receiving a first allogeneic HCT from 2013-2017, were included. Patients were stratified according to mELN genetic classification in three distinct groups: favorable (Fav), intermediate (IM) and adverse (Adv). Clinical outcomes following HCT were compared among groups after adjusting for significant patient, disease, and transplant-related variables. The primary endpoint was disease free survival (DFS). Secondary endpoints were overall survival (OS), non-relapse mortality (NRM), cumulative incidence of relapse, acute GVHD, and chronic GVHD. Cox proportional hazard models were used to compare endpoints among mELN risk groups, age groups and genetic subsets within the adverse-risk group. Results: Demographic characteristics are summarized in Table 1. 2289 patients (Fav, n=181; IM n=1185; and Adv n=923) met the inclusion criteria. Median follow-up for survivors was 35 months. Importantly, 41% of transplant recipients (n=936) were &gt;60 years, 76% (n=1743) had de novo AML and 48% (n=1111) received a myeloablative conditioning regimen. Univariate analysis (UVA) demonstrated significant differences in 2-year OS (Fav: 67.7%, IM: 64.9% and Adv: 53.9%; p&lt;0.001); DFS (Fav: 57.8%, IM: 55.5% and Adv: 45.3; p&lt;0.001) and relapse (Fav: 28%, IM: 27.5% and Adv: 37.5%; p&lt;0.001) There were no significant differences in NRM (p=0.467) or the incidence of acute (p=0.423) and chronic GVHD (p=0.442) among mELN groups. Initial multivariate analysis (MVA) of mELN risk groups indicated that there was no significant difference in clinical outcomes between the Fav and IM risk groups. Thus, these groups were combined for subsequent analyses. Adv risk (vs. Fav/IM) led to significantly worse OS (HR 1.39 [1.24-1.57] p=&lt;0.001), DFS (HR 1.32 [1.18-1.48] p=&lt;0.001), and relapse (HR 1.42 [1.23-1.63] p=&lt;0.001) (Table 2). This mELN classification effectively stratified both younger (&lt;60 y/o) and older (&gt;60 y/o) patients for OS (Adv vs. Fav/IM HR for &lt;60: 1.43 [1.21-1.69] p&lt;0.001; &gt;60: 1.44 [1.21-1.72] p&lt;0.001), DFS (Adv HR for &lt;60: 1.31 [1.12-1.53] p&lt;0.001; &gt;60: 1.41 [1.19-1.66] p&lt;0.001) and relapse (Adv HR for &lt;60: 1.44 [1.20-1.74] p&lt;0.001, &gt;60: 1.42 [1.15-1.76] p=0.001). NRM was higher for older patients (&gt;60 y/o) in both the Fav/IM (HR 1.40 [1.10-1.78] p=0.007) and Adv-risk cohorts (HR 1.59 [1.18-2.13] p=0.002). Genetic subset comparisons within the adverse-risk group showed that patients carrying monosomy 5, del(5q) or monosomy 7 had inferior 2-year OS (42.6%, p&lt;0.001) and DFS (35.2%, p&lt;0.001), as well as higher rates of relapse (45%, p= 0.002) when compared to other patients within the Adv-risk cohort (OS 60%, DFS 50.8%, relapse 33.5%). Conclusion: Stratification using mELN criteria resulted in clear prognostic separation of OS, DFS and relapse in this large cohort of AML patients undergoing allogeneic HCT. While Fav and IM groups had similar OS, DFS and relapse rates; patients in the Adv risk group had the highest risk of relapse and inferior DFS/OS. However, the majority of patients in all cohorts had favorable outcomes for HCT in CR1. Our findings confirm the value of a combined genetic prognostic model in the AML HCT setting and justify the use of this stratification system in future HCT trials. Correlation of genetic subtypes with other important transplant variables, such as conditioning intensity and pre-transplant MRD status deserves further evaluation. There remains a subset of Adv-risk patients for which post-transplant outcomes continue to be poor, even when transplanted in CR1. Novel peri-transplant pre-emptive/therapeutic strategies are urgently needed for this high-risk cohort. Disclosures de Lima: Celgene: Research Funding; Pfizer: Other: Personal fees, advisory board, Research Funding; BMS: Other: Personal Fees, advisory board; Incyte: Other: Personal Fees, advisory board; Kadmon: Other: Personal Fees, Advisory board. Komanduri:Kiadis: Consultancy; Takeda: Consultancy; Celgene: Consultancy; Atara: Consultancy, Membership on an entity's Board of Directors or advisory committees; Adaptimmune: Membership on an entity's Board of Directors or advisory committees; Novartis: Consultancy, Membership on an entity's Board of Directors or advisory committees; Kite/Gilead: Consultancy, Membership on an entity's Board of Directors or advisory committees. Kebriaei:Ziopharm: Other: Research Support; Jazz: Consultancy; Kite: Other: Served on advisory board; Amgen: Other: Research Support; Novartis: Other: Served on advisory board; Pfizer: Other: Served on advisory board. Bhatt:National Marrow Donor Program: Research Funding; Rigel Pharmaceuticals: Other; Jazz: Research Funding; Agios: Other: Personal Fees; Incyte: Other: Personal Fees, Research Funding; Takeda: Other: Personal Fees; Partner Therapeutics: Other: Personal Fees; Pfizer: Other, Research Funding; CSL Behring: Other; Tolero Pharmaceuticals: Research Funding; Oncoceutics: Other: Drug support for a trial; Partnership for health analytic research, LLC: Other: Personal Fees; Omeros: Other: Personal Fees; Abbvie: Other: Personal Fees, Research Funding. Saad:Orcabio: Other: research support; Kadmon: Other: research support; Amgen: Other: research support; Incyte Pharmaceuticals: Other: Personal Fees; Magenta Therapeutics: Other: Personal Fees. Copelan:Amgen: Membership on an entity's Board of Directors or advisory committees. Defilipp:Incyte: Research Funding; Regimmune: Research Funding; Syndax Pharmaceuticals: Consultancy. Nishihori:Karyopharm: Other: Research support to institution; Novartis: Other: Research support to institution. Weisdorf:Incyte: Research Funding; FATE Therapeutics: Consultancy.

ObjectivesTo determine whether renal transplant diagnoses substantially change when 2 biopsy sites are chosen and whether contrast‐enhanced ultrasound (CEUS) has value for targeting the second site.MethodsWe prospectively enrolled 40 patients undergoing ultrasound‐guided renal transplant biopsy within 2 years of transplant: 20, surveillance; and 20, for cause. A CEUS examination was performed to identify cortical regions with subjectively altered flow. One biopsy was performed at the operator‐preferred (primary) site regardless of CEUS findings. Another biopsy was done at a second location, either targeted to an area in which CEUS perfusion findings differed from the primary site (targeted) or at a random location (secondary) if no other area differed. Specimens were randomly labeled A or B; pathologists were blinded to the CEUS result and biopsy location. Location‐specific CEUS assessments were recorded. Pathologic results were compared, including acute and chronic Banff scores and any new findings from the targeted or secondary biopsy.ResultsForty patients were enrolled between January 2016 and December 2018. No location‐specific pathologic differences correlated with differences in CEUS assessments. The second biopsy provided additional information that changed management in 4 of 40 patients (10.0% [95% confidence interval, 2.8%–23.7%]). Major bleeding complications occurred in 3 of 40 (7.5%) patients.ConclusionsContrast‐enhanced ultrasound targeting was not useful. Major bleeding complications were higher than expected, possibly due to the additional biopsy away from the operator‐preferred location. Obtaining a second renal transplant biopsy from a substantially different area than the initial operator‐preferred location provided additional clinically useful information in 10% of patients.