Alternative Score Research Articles

Study design: Development of an advanced machine learning algorithm for the early diagnosis of Gaucher disease using real-world data

Background: Gaucher disease (GD) is a rare, autosomal recessive condition, characterized by deficiency of the lysosomal enzyme β-glucocerebrosidase. The main disease features are anemia, thrombocytopenia, hepato-splenomegaly and bone infarction, osteonecrosis, and pathological fractures. However, diagnosis of GD can be challenging, especially for non-specialists, owing to wide variability in age at presentation, non-specific features, severity and type of clinical manifestations, and lack of awareness of the early signs and symptoms of the disease. Delayed and misdiagnosis of GD may lead to irreversible bone disease, severe growth retardation, and high risk of bleeding; in rare cases, misdiagnosis may be life-threatening. Developing a system for early and accurate diagnosis of GD is thus an essential unmet need. The development of an algorithm for early diagnosis of patients with rare diseases such as GD may help reduce delays in diagnosis and enable prompt, appropriate initiation of therapy, earlier decision-making, prevent potentially irreversible morbidities and unnecessary tests (some invasive), reduce anxiety, and facilitate genetic counseling. This study aims to develop a predictive model for the accurate diagnosis of GD using machine learning based on real-world clinical data. Methods: This study will be comprised of three parts. Part 1, a retrospective observational database analysis, will use data from the electronic patient database of the Maccabi Healthcare Service (MHS), the second largest Health Maintenance Organization in Israel. The MHS includes 2.2 million health records from 25% of the Israeli population. Clinical records have been fully computerized for >20 years and are fully integrated with automated central laboratory, digitized imaging and pharmacy purchase data. Patients with confirmed GD who have been enrolled in the MHS health plan for ≥1 year will be eligible for inclusion, with approximately 250 patients with GD expected to be enrolled. Using MHS data from patients with GD, the Gaucher Earlier Diagnosis Consensus (GED-C) scoring system, developed by a consensus panel using Delphi methodology on the signs and co-variables that may be important for the diagnosis of GD, will be evaluated and compared with alternative scores developed directly from clinical data based on supervised machine learning. In Part 2, a clinical study, the best performing modeled scores from Part 1 will be applied to the MHS database to identify individuals who may have undiagnosed GD (‘GD suspects‘). Samples for diagnostic testing (using a specific and sensitive biomarker (glucosylsphingosine, lyso-Gb1) followed by beta-glucocerebrosidase (GBA) genotyping for positive samples) will be collected from MHS biobank (for individuals who have consented). Individuals not participating in the biobank will be asked to provide a sample. This part of the study will evaluate the predictive value of the modeled scores, and assess the sensitivity and specificity of the model for the diagnosis of new patients with GD. In Part 3, analysis of data from newly diagnosed patients identified in Part 2 will be used to develop machine learning models for the diagnosis of GD (Figure 1). Signs and co-variables included in the GED-C score will be used, eliminating features that are non-informative. Features will be quantitative where possible, and interaction terms will be added for age of onset and trend for key features. A number of methods will be developed, with the best performing, based on its precision at a given sensitivity level, being selected as the final model. External validation of the best identified model is planned, to ensure unbiased estimate of the model's accuracy. Discussion: The main goal of the study is to develop an algorithm to help detect patients with GD, independent of physicians' ability to recognize signs and symptoms, using the application of machine learning to data from a large health database. The study is expected to result in a practical tool that will alert physicians to the possibility of GD. The resulting model will also improve our understanding of GD based on the relative importance of features for GD prediction. Such tools will have a positive impact on patient care and quality of life and on healthcare costs and may lead to a change in approach for diagnosing rare diseases. Download : Download high-res image (323KB) Download : Download full-size image Disclosures Revel-Vilk: Takeda: Honoraria; sanofi-Genzyme: Honoraria; Pfizer: Honoraria. Chodick: Novartis Pharma AG: Other: Institutional grant. Gadir: Takeda: Current Employment.

Perioperative acinar cell count method works well in the prediction of postoperative pancreatic fistula and other postoperative complications after pancreaticoduodenectomy

BackgroundEarlier we have shown that high frequency of acinar cells in the pancreatic transsection line predicts postoperative pancreatic fistula after pancreaticoduodenectomy (PD). Acinar cell count method (ACM) is fast to perform during operation. In this study our aim was to validate the accuracy of ACM to compare it with other published risk prediction methods. Methods87 patients who underwent PD without any trial including perioperative medications were collected from a single hospital. Data on demographics, surgical details, postoperative complications clinically relevant pancreatic fistulae (CR-POPF) and clinically relevant Clavien-Dindo complications (CR-CDC) were registered. Thirteen previously published risk prediction methods were included in the comparison, such as pancreatic duct diameter, palpable texture of pancreas, Braga score (BC), Fistula Risk Score, Modified Fistula Risk Score, Alternative Fistula Risk Score and multiple radiological parameters. ROC-curves were calculated to compare sensitivity and specificity for identifying high risk patients for CR-POPF and CR-CDC. ResultsThe three most accurate risk prediction methods for CR-POPF were ACM (sensitivity 88.9%, specificity 52.6%; p = 0.043), BC (87.5%, 56.6%; p = 0.039) and visceral fat area to subcutaneous fat area ratio (75.5%, 80.0%; p = 0.032). In predicting CR-CDC the three most accurate methods were ACM (73.9%, 56.2%; p = 0.033), BC (68.4%, 59.5%; p = 0.036) and TPAI (78.3%, 41.7%; p = 0.012). ConclusionACM was shown to be as good as the more complicated risk scoring methods in the prediction of CR-POPF. It was good also in predicting all clinically relevant complications. ACM is easy to use during operation and can be recommended as a routine risk prediction method.

External validation of four different models of prediction the clinically relevant pancreatic fistula after laparoscopic pancreatoduodenectomy

Introduction: Different models were proposed to predict clinically relevant postoperative pancreatic fistula (CR-POPF) development after open pancreatoduodenectomy. Among them are the Fistula Risk Score (FRS), Alternative Fistula Risk Score (a-FRS), Updated Alternative Fistula Risk Score (ua-FRS), Japanese Fistula Risk Score (JFRS). External validations of these scales applying to laparoscopic procedure should be performed. Purpose: Is to perform external validations of different scales of prediction CR-POPF after LPDE. Methods: From 2007 to 2018, 290 LPDEs were performed at single institution. The individual risk scores according different scales were calculated for each patient, and clinical outcomes were evaluated. Receiver operator curve analysis was performed to judge model validity Results: Out of 290 patients, 60 (20,7%) developed CR-POPF (15,9 % grade B, 4,8 % grade C). The incidences of CR-POPF in patients classified as negligible risk, low risk, intermediate risk, and high-risk by original-FRS were 4%, 8.2%, 26%, and 45%, respectively. The area under the ROC curve (AUC) was 0.745 for a-FRS vs. 0.70 for original-FRS. The incidences of CR-POPF in patients classified as low risk, moderate risk, and high-risk by a-FRS were 7,5%, 23,9%, and 44%, respectively. The AUC was 0.748. In patients who developed CR-POPF the mean ua-FRS score was 43,5%±21,9 vs 23,9%±18,7 for patients who did not suffer from CR-POPF (р< 0,0001). The AUC for ua-FRS was 0,772. The AUC of J-FRS, which was originally developed for the prediction of grade C POPF, was 0,692, showing low predictive value of this scale for LPDE. Conclusion: The ua-FRS appears to be the most accurate and convenient tool for predicting occurrence of CR-POPF after LPDE.

Design and Regulation of Balancing Power Auctions – An Integrated Market Model Approach

We present an integrated market model which considers the dependencies between the wholesale market and the highly regulated balancing power markets. This fosters the understanding of the mechanisms of these markets and, thus, allows the evaluation of the designs of these markets and their interplay. In contrast to existing literature, in our model the prices on the different markets are interdependent and endogenously determined, which also applies to the switch from inframarginal suppliers to extramarginal suppliers. Linked to this, the implementation of a specific assignment of the suppliers to the different markets is according to their production costs and their ability to provide balancing power. We prove the existence of a market equilibrium, analyze its outcome and contrast this with German market data. Based on this model, we assess design changes, partly stipulated by recent European regulation. This includes uniform pricing as a common settlement rule (effect: no truthful bidding in general), standardized prequalification criteria (promising measure for cost reduction), market flexibilization via “free energy bids” (no increased competition) and the alternative score “mixed-price rule” (no effect on the equilibrium).

Protective Peritoneal Patch for Arteries during Pancreatoduodenectomy: Good Value for Money

Purpose This study proposes and details a simple and inexpensive protective technique of wrapping the hepatic and gastroduodenal artery stumps with a peritoneal patch during pancreatoduodenectomy (PD) in order to decrease post-pancreatectomy hemorrhage (PPH). Methods Among the 85 patients who underwent PD between July 2020 and March 2021, 16 patients with high-risk pancreatic anastomosis received a peritoneal patch. The Updated Alternative Fistula Risk Score (ua-FRS) was calculated. Post-operative pancreatic fistula (POPF) and PPH were diagnosed and graded according to the International Study Group of Pancreatic Surgery. The mortality rate was calculated up to 90 days after PD. Results The mean ua-FRS of the 16 patients was 43% (range: 21-63%). Among them, 6 (38%) experienced clinically relevant-POPF, and a PPH was observed in two patients (13%). In these two patients who required re-intervention, the peritoneal patch was remarkably intact, and neither the gastroduodenal stump nor hepatic artery was involved. None of the patients experienced 90-day mortality. Conclusion Although the outcomes are encouraging, the evaluation of a larger series to assess the effectiveness of the peritoneal protective patch for arteries in a high-risk pancreatic anastomosis is ongoing.

Alternative fistula risk score is a better predictor of ideal outcome in patients undergoing minimally invasive pancreaticoduodenectomy

Background: The benefit of minimally-invasive (MIPD) versus open pancreaticoduodenectomy (OPD) to patients is controversial. The major driver of morbidity and mortality after pancreaticoduodenectomy is postoperative pancreatic-fistula (POPF). The alternative-fistula risk score (aFRS) is a validated prognostic tool that predicts patients' risk of POPF using three variables (pancreatic duct size, gland texture, and patient BMI). We hypothesized that patients who are not at high risk for POPF benefit most from MIPD. Methods: Patients undergoing pancreaticoduodenectomy were prospectively followed for 40 months. Perioperative and pathologic covariates and outcomes were compared. Patients were categorized as either aFRS-high risk (POPF-risk >20%) or aFRS-low/intermediate-risk (POPF-risk ≤20%). The ideal-outcome (IO) was defined as shown in Figure-1. Multivariable logistic regression was used to test for independent-associations with IO. Results: Out of 312 patients, 212 (83.7%) underwent OPD and 51 (16.4%) underwent MIPD. MIPD patients had significantly longer overall operative time (462.8 minvs.378.5 min,p< 0.001), reduced intraoperative blood loss (280.5 mlvs.436.0 ml,p=0.001), and a decreased rate of 90-day readmission (15.7%vs.30.7%,p=0.030) . MIPD patients were significantly more likely to be aFRS-high risk (56.9%vs.40.2%,p=0.028). IO was significantly more frequent in aFRS-low/intermediate-risk patients undergoing MIPD compared to aFRS-low/intermediate-risk patients undergoing OPD, aFRS-high risk patients undergoing either MIPD or OPD (40.9%vs15.4%v6.9% vs14.3%,p=0.007). In multivariate analysis, MIPD in aFRS-low/intermediate-risk patients was independently-associated with an increased likelihood of IO (OR= 4.09, p=0.012). Conclusions: Patients who are not at high-risk for POPF are most likely to benefit from MIPD. The aFRS could be a useful tool to aid the surgeon experience and expertise in selecting patients for MIPD.

Risk Factors for Development of Clinically Relevant Pancreatic Fistula after Laparoscopic Pancreatoduodenectomy in High-risk Patients According to the ua-FRS

Introduction: Updated Alternative Fistula Risk Score (ua-FRS) is the accurate tool to predict clinically relevant postoperative pancreatic fistula (CR-POPF) after laparoscopic pancreatoduodenectomy (LPDE). While the risk factors of development of the CR-POPF are well studied, parameters leading to formation CR-POPF in patients of high-risk group are yet to be defined. Objective: To define the risk factors of development of the CR-POPF in patients of high-risk group according to the ua-FRS. Methods: From 2007 to 2018, 290 LPDEs were performed at single institution. The individual risk scores according ua-FRS were calculated for each patient, and clinical outcomes were evaluated. A high-risk group of CR-POPF formation was identified. Risk factors for CR-POPF were evaluated by univariate and multivariate regression analysis. Results: Out of 290 patients, 110 (38%) were allocated to the high-risk group. CR-POPF rate was 39% in this group, whereas it was 21,2% and 9,2% in the whole group and low-risk group, respectively. Univariate analysis revealed that the performing the procedure by the lowest-experienced surgeon and the diagnosis other than pancreatic adenocarcinoma were independent risk factors for CR-POPF formation. The multivariate regression analysis revealed, that Charlson comorbidity index (CCI) higher than 3 is associated with CR-POPF (OR = 2,205, CI 95%, -0,026-0,436). Conclusion: Performing the LPDE by the low-experienced surgeon, diagnosis other than pancreatic adenocarcinoma, CCI higher than 3 are associated with CR-POPF development after LPDE in high-risk patient group.

High-risk Pancreatic Anastomosis vs. Total Pancreatectomy after Pancreatoduodenectomy: Postoperative Outcomes and Quality of Life Analysis

Purpose: Outcomes of pancreaticoduodenectomy (PD) and total pancreatectomy (TP) in patients at high-risk for postoperative pancreatic fistula (POPF) have never been compared. The aim of this study was to evaluate TP as an alternative to high-risk pancreaticoduodenectomy (HR-PD). Methods: All patients who underwent PD or TP between July 2017 and December 2019 were identified. HR-PD was defined according to the alternative Fistula Risk Score. Postoperative outcomes (primary endpoint), pancreatic insufficiency and quality of life after 12 months of follow-up (QoL) were compared between HR-PD or planned PD intraoperatively converted to TP (C-TP). Results: A total of 566 patients underwent PD and 136 underwent TP during the study period. One hundred one (18%) PD patients underwent HR-PD, while 86 (63%) TP patients underwent C-TP. Postoperatively, the patients in the C-TP group exhibited lower rates of post-pancreatectomy hemorrhage (15% vs 28%), delayed gastric emptying (16% vs 34%), sepsis (10% vs 31%), and Clavien-Dindo ≥3 morbidity (19% vs 31%) and had shorter median lengths of hospital stay (10 vs 21 days) (all p< 0.05). The rate of POPF in the HR-PD group was 39%. Mortality was comparable between the two groups (3% vs 4%). Although general, cancer- and pancreas-specific QoL were comparable between the HR-PD and C-TP groups, endocrine and exocrine insufficiency occurred in all the C-TP patients, compared to only 13% and 63% of the HR-PD patients respectively, and C-TP patients had worse diabetes-specific QoL. Conclusions: C-TP may be considered rather than HR-PD only in few selected cases and after adequate counselling.

Somatostatin Analogues for the Prevention of Pancreatic Fistula after Open Pancreatoduodenectomy: A Nationwide Analysis

Purpose: Somatostatin analogues (SA) are varyingly used to try to prevent development of a pancreatic postoperative fistula (POPF) after pancreatoduodenectomy. Some centers do not use SA, some use it only in high-risk patients, and others in all patients. The aim of this study was to evaluate the effect of different protocols for SA use on the incidence of POPF after pancreatoduodenectomy in the Netherlands. Method: The Dutch Pancreatic Cancer Audit (DPCA) is a national, mandatory audit that registers all patients after pancreatoduodenectomy. Consecutive patients undergoing elective open pancreatoduodenectomy for all indications were included from the DPCA (2014-2017). Patients were divided into six groups depending on the SA protocol: no SA, octreotide, lanreotide, pasireotide, octreotide only in high risk patients, and lanreotide only in high risk patients. Primary endpoint was POPF (grade B and C). The updated alternative Fistula Risk Score (uaFRS) was used to analyze the occurrence of clinically relevant POPF across various risk scenarios (low-intermediate risk (<20%) and high risk (≥20%)). Results: Overall, 1992 patients were included with a POPF rate of 13.1%. POPF occurred less often when octreotide was administered in high risk patients only (9.4%) or if a lanreotide protocol was used (10.0%) compared to other protocols (varying from 12.7-19.1%, p=0.001). Sub-analysis in 577 patients (56.7%) with a high uaFRS showed a significantly lower rate of POPF when treated with lanreotide (10.0%) compared to the other protocols (varying from 18.9%-27.4%) (p=<0.001). In the 440 patients (43.3%) with a low-intermediate uaFRS the rate of POPF did not differ significantly between the six protocols. Multivariate regression analysis, controlling for male sex, BMI, soft pancreatic tissue, a pancreatic duct smaller than 3mm, ASA score III/IV, neuroendocrine tumor pathology, demonstrated lanreotide to be significantly associated with a low OR for POPF (OR 0.387, 95% CI 0.180-0.834) compared to no SA use. Conclusion: This nationwide retrospective study suggests that the use of lanreotide in patients undergoing pancreatoduodenectomy is beneficial in reducing the rate of POPF. However, perioperative mortality is unaffected. Since SA appear not to be beneficial in low-intermediate risk patients, the use of lanreotide may be reserved for patients at high risk of POPF.

Assignment Mechanisms with Public Preferences and Independent Types

The literature on delegation considers the problem of an uninformed decision maker and an informed but biased agent. I extend that analyis to the case of multiple agents under two assumptions: independent private information and public preferences. In the optimal mechanism, agents assign points to the various alternatives, which then get mapped into scores, so that the alternative with the largest score wins. Each alternative's score is the sum of points received plus an extra term that is larger when the agents who have a strong preference for that alternative assign points to the alternatives they like less.

Organic food consumption is associated with inflammatory biomarkers among older adults.

The association between organic food consumption and biomarkers of inflammation, C-reactive protein (CRP) and cystatin C (CysC) was explored in this cross-sectional analysis of older adults. Dietary data and organic food consumption was collected in 2013 from a FFQ. Alternative Mediterranean diet score (A-MedDiet) was calculated as a measure of healthy eating. Biomarkers CRP and CysC were collected in serum or plasma in 2016. We used linear regression models to assess the associations between organic food consumption and CRP and CysC. This cross-sectional analysis uses data from the nationally representative, longitudinal panel study of Americans over 50, the Health and Retirement Study. The mean age of the analytic sample (n 3815) was 64·3 (se 0·3) years with 54·4 % being female. Log CRP and log CysC were inversely associated with consuming organic food after adjusting for potential confounders (CRP: β = -0·096, 95 % CI 0·159, -0·033; CysC: β = -0·033, 95 % CI -0·051, -0·015). Log CRP maintained statistical significance (β = -0·080; 95 % CI -0·144, -0·016) after additional adjustments for the A-MedDiet, while log CysC lost statistical significance (β = -0·019; 95 % CI -0·039, 0·000). The association between organic food consumption and log CRP was driven primarily by milk, fruit, vegetables and cereals, while log CysC was primarily driven by milk, eggs and meat after adjustments for A-MedDiet. These findings support the hypothesis that organic food consumption is inversely associated with biomarkers of inflammation CRP and CysC, although residual confounding by healthy eating and socioeconomic status cannot be ruled out.

Factors associated with abnormal glucose readings in a pharmacy-led community health fair using the ADA risk assessment tool

ObjectivesThe objective of this study was to evaluate whether the American Diabetes Association (ADA) risk tool correctly identified high-risk patients with abnormal glucose readings requiring referral to a health care provider in an underserved population at a large student pharmacist–led health fair. In addition, the association of abnormal glucose readings compared with alternative ADA risk score cut point values and other collected variables was evaluated. MethodsThis was a retrospective, cross-sectional study using deidentified data from a large student pharmacist–led health fair. ResultsA total of 35 of 188 (19%) patients were considered high risk per the ADA risk tool, and 11 of those 35 (31%) had abnormal glucose results. After controlling for ADA risk score, no additional collected clinical variables were independently associated with abnormal glucose results. Although the ADA cut point associated with an abnormal glucose screening with the highest area under the curve was greater than or equal to 4, a cut point of 3 or greater resulted in a sensitivity of 91.2%. ConclusionThe optimal method to identify patients who are at risk for an abnormal glucose screening is the ADA risk tool compared with the individual components of the tool or other evaluated risk factors. We suggest using an ADA risk cut point of greater than or equal to 3 instead of greater than or equal to 5 to identify patients likely to have abnormal glucose results in the health fair setting as this would greatly increase the chance of identifying patients who would need to be referred to their primary care provider for diagnostic testing.

External Validation of the HLH-2004 Diagnostic Criteria and H-Score for Diagnosis of Hemophagocytic Lymphohistiocytosis in Adults

Introduction: Hemophagocytic lymphohistiocytosis (HLH) is a hyperinflammatory syndrome associated with multi-organ failure and death. Diagnosis in adult patients is currently based on the HLH-2004 diagnostic criteria; however, these criteria were developed for pediatric HLH and have not been formally validated in adults. An alternative diagnostic score, the H-score, was developed for adults with reactive HLH. There have been few external validation studies comparing the diagnostic accuracy between the HLH-2004 criteria and H-score, mostly in critically ill patients. In this external validation study, we aimed to compare the discriminatory power of these two diagnostic criteria in predicting HLH in a multicenter cohort of adults with suspected HLH. Methods: We identified all adult inpatients (≥18 years) with an International Classification of Diseases (ICD) code for HLH in the province of Alberta, Canada from January 1999 to December 2019. Following independent chart review by two reviewers, cases were classified as positive, negative, or indeterminate cases of HLH. The HLH-2004 diagnostic criteria and H-score were determined for each case. The following performance characteristics of the diagnostic scoring systems were calculated: sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). C-statistic was calculated from the receiver operator curve (ROC) analysis. Results: Data on 101 patients were collected. The median age was 46 years (range 18-88), 66 (65%) were male. Active malignancy within 6 months of presentation, infections, and autoimmune diseases were present in 31%, 36%, and 18%, respectively. Of the 96 patients who underwent bone marrow aspirate and biopsy, hemophagocytosis was present in 79 (82%). Using chart review as the gold standard, 89 (88%) patients were determined to have HLH, whereas 5 (5%) and 7 (7%) were reclassified as negative and indeterminate cases, respectively. 83 (82%) patients met ≥5 of the 8 HLH-2004 diagnostic criteria; using the HLH-2004 criteria missed 8 cases of HLH. The sensitivity and specificity of HLH-2004 criteria was 91.0% and 83.3%, respectively, with a PPV of 97.6% and an NPV of 55.6%. The median overall H-score was 238 (range 129-337). An H-Score cut-off of &amp;gt;169 (established as the optimal cut-off from the original derivation and validation paper) predicted HLH with a sensitivity of 95.5%, specificity of 16.7%, PPV of 89.5%, and NPV of 33.3%. The discriminatory power improved with an H-Score cut-off &amp;gt;210, with a sensitivity of 86.5%, specificity of 66.7%, PPV of 95.1%, and NPV of 40.0%. The c-statistics for the HLH-2004 criteria and the H-Score were 0.872 and 0.737, respectively. In the malignant subgroup, all patients fulfilled the HLH-2004 criteria and all were deemed to have HLH upon chart review. All but one had an H-Score &amp;gt;169, with a median H-Score of 238 (range 168-304) in the malignant group. Conclusion: In our highly selected patient population with suspected HLH, the HLH-2004 had better discriminatory power than the H-Score, with a high sensitivity and specificity and excellent c-statistic &amp;gt;0.80. An H-Score cut-off of &amp;gt;169 was highly sensitive but non-specific in identifying adults with HLH. The validity and applicability of our study is limited by a high prevalence of chart-confirmed HLH in our sample. We are completing ongoing data collection to expand our validation study in a cohort of adults with hyperferritinemia who underwent bone marrow evaluation and/or sCD25 testing. Disclosures Sun: Sanofi: Other: Advisory board; Octapharma: Other: Advisory board; Octapharma: Research Funding; Novo Nordisk: Other: Advisory board; Pfizer: Other: Advisory board.

Study Design: Development of an Advanced Machine Learning Algorithm for the Early Diagnosis of Gaucher Disease Using Real-World Data

Background: Gaucher disease (GD) is a rare, autosomal recessive condition, characterized by deficiency of the lysosomal enzyme β-glucocerebrosidase. The main disease features are anemia, thrombocytopenia, hepato-splenomegaly and bone infarction, osteonecrosis, and pathological fractures. However, diagnosis of GD can be challenging, especially for non-specialists, owing to wide variability in age at presentation, non-specific features, severity and type of clinical manifestations, and lack of awareness of the early signs and symptoms of the disease. Delayed and misdiagnosis of GD may lead to irreversible bone disease, severe growth retardation, and high risk of bleeding; in rare cases, misdiagnosis may be life-threatening. Developing a system for early and accurate diagnosis of GD is thus an essential unmet need. The development of an algorithm for early diagnosis of patients with rare diseases such as GD may help reduce delays in diagnosis and enable prompt, appropriate initiation of therapy, earlier decision-making, prevent potentially irreversible morbidities and unnecessary tests (some invasive), reduce anxiety, and facilitate genetic counseling. This study aims to develop a predictive model for the accurate diagnosis of GD using machine learning based on real-world clinical data. Methods: This study will be comprised of three parts. Part 1, a retrospective observational database analysis, will use data from the electronic patient database of the Maccabi Healthcare Service (MHS), the second largest Health Maintenance Organization in Israel. The MHS includes 2.2 million health records from 25% of the Israeli population. Clinical records have been fully computerized for &amp;gt;20 years and are fully integrated with automated central laboratory, digitized imaging and pharmacy purchase data. Patients with confirmed GD who have been enrolled in the MHS health plan for ≥1 year will be eligible for inclusion, with approximately 250 patients with GD expected to be enrolled. Using MHS data from patients with GD, the Gaucher Earlier Diagnosis Consensus (GED-C) scoring system, developed by a consensus panel using Delphi methodology on the signs and co-variables that may be important for the diagnosis of GD, will be evaluated and compared with alternative scores developed directly from clinical data based on supervised machine learning. In Part 2, a clinical study, the best performing modeled scores from Part 1 will be applied to the MHS database to identify individuals who may have undiagnosed GD ('GD suspects'). Samples for diagnostic testing (using a specific and sensitive biomarker (glucosylsphingosine, lyso-Gb1) followed by beta-glucocerebrosidase (GBA) genotyping for positive samples) will be collected from MHS biobank (for individuals who have consented). Individuals not participating in the biobank will be asked to provide a sample. This part of the study will evaluate the predictive value of the modeled scores, and assess the sensitivity and specificity of the model for the diagnosis of new patients with GD. In Part 3, analysis of data from newly diagnosed patients identified in Part 2 will be used to develop machine learning models for the diagnosis of GD (Figure 1). Signs and co-variables included in the GED-C score will be used, eliminating features that are non-informative. Features will be quantitative where possible, and interaction terms will be added for age of onset and trend for key features. A number of methods will be developed, with the best performing, based on its precision at a given sensitivity level, being selected as the final model. External validation of the best identified model is planned, to ensure unbiased estimate of the model's accuracy. Discussion: The main goal of the study is to develop an algorithm to help detect patients with GD, independent of physicians' ability to recognize signs and symptoms, using the application of machine learning to data from a large health database. The study is expected to result in a practical tool that will alert physicians to the possibility of GD. The resulting model will also improve our understanding of GD based on the relative importance of features for GD prediction. Such tools will have a positive impact on patient care and quality of life and on healthcare costs and may lead to a change in approach for diagnosing rare diseases. Disclosures Revel-Vilk: Takeda: Honoraria; sanofi-Genzyme: Honoraria; Pfizer: Honoraria. Chodick:Novartis Pharma AG: Other: Institutional grant. Gadir:Takeda: Current Employment.

Abstract PO-30: Is it more explanatory to integrate the leukocyte/lymphocyte ratio (LLR) and prognostic nutritional index (PNI) to international prognostic systems (IPS) in cases with Hodgkin lymphoma (HL)?

Abstract Aim: To determine the power of the International Prognostic Scoring systems (IPS-7 and IPS-3) and to obtain more explanatory indexes by integrating LLR and PNI as prognostic indicators to IPS in cases with HL. Patients and Methods: 1,060 patients with HL were included in this study from 8 centers in Turkey. IPS-7 using seven parameters (age, sex, stage hemoglobin, albumin, lymphocyte count, and white cell count) and the modified IPS-3 score with three parameters (stage, age, and hemoglobin level) were calculated. An alternative IPS-4 score with four parameters (stage, age, hemoglobin level, and LLR or PNI) was also calculated. LLR and PNI were integrated to IPSs. Statistical Analyses: Chi square test or Student t test was used to compare the groups. The cut-off points of this population for the LLR and PNI were evaluated using AUC (area under ROC curve) statistics and median values of the distributions of the variables. Possible alternative cut-off points for the LLR and PNI were evaluated using AUC statistics and median values of the distributions of the variables. For LLR, 5.7 was detected as cut-off point based on ROC analysis and 5.1 based on median. Cut-off for PNI was 45.2 based on ROC analysis and 47.5 based on median. The Kaplan-Meier method and Cox proportional regression model were used to estimate the mean-median overall survival (OS), failure-free survival (FFS) rates, and hazard ratios (HRs). Log-rank test was used to compare the survival distributions between groups. The prognostic ability of parameters was evaluated for OS and FFS in both univariate and multivariate Cox regression models. p value &amp;lt; 0.05 was considered as significant. The analyses were performed using the statistical package SPSS v 22.0. Results: Female/male ratio was 396/664; mean age was 37.6±15.8 (15-88). About 75% of the patients had stage II-III disease and two thirds of the patient had nodular sclerosis subtype. All factors of IPS-7 and IPS-3 scoring systems were found to be significant factors that related with OS and FFS according to univariate analyses. PNI and LLR were also found to be prognostic factors for OS and FFS. PNI ≤45.2 and LLR ≥5.8 were found to be poor prognostic indicators. Among 7 factors of IPS-7, gender and albumin were not found to be significant according to multivariate Cox regression model. Hence albumin and gender were excluded from the models and PNI and LLR as prognostic factors were added to obtain easier and more explanatory indexes to the model, respectively. Two different Cox regression models were obtained for OS and FFS. Model 1 showed LLR ≥5.8 as the highest risk for OS (OR: 2.7) and Hb &amp;lt;10.5g/dL as the highest risk for FFS (OR: 2.3). Model 2 showed PNI ≤45.2 as the highest risk for OS (OR: 3.2) and Hb &amp;lt;10.5g/dL as the highest risk for FFS (OR: 2.9). Goodness of fit and agreement between IPS-7, IPS-3 groups and IPS-4 groups were found to be acceptable. Conclusion: IPS-4 score obtained by integrating LLR or PNI to IPS systems can be used as an explanatory index for prognostic indicator in HL. Citation Format: Semra Paydas, Sahin Lacin, Mutlu Dogan, Ibrahim Barista, Birol Yildiz, Gulsah Seydaoglu, Nuri Karadurmuş, Sinem Civriz, Muhammed Ali Kaplan, Munci Yagci, Emel Gurkan, Vehbi Ercolak. Is it more explanatory to integrate the leukocyte/lymphocyte ratio (LLR) and prognostic nutritional index (PNI) to international prognostic systems (IPS) in cases with Hodgkin lymphoma (HL)? [abstract]. In: Proceedings of the AACR Virtual Meeting: Advances in Malignant Lymphoma; 2020 Aug 17-19. Philadelphia (PA): AACR; Blood Cancer Discov 2020;1(3_Suppl):Abstract nr PO-30.

Perioperative Acinar Cell Count Method Works Well in the Prediction of Postoperative Pancreatic Fistula and Other Postoperative Complications After Pancreaticoduodenectomy

Earlier we have shown that high frequency of acinar cells in the pancreatic transsection line predicts postoperative pancreatic fistula after pancreaticoduodenectomy (PD). Acinar cell count method (ACM) is fast to perform during operation. In this study our aim was to validate the accuracy of ACM to compare it with other published risk prediction methods. 87 patients who underwent PD without any trial including perioperative medications were collected from a single hospital. Data on demographics, surgical details, postoperative complications clinically relevant pancreatic fistulae (CR-POPF) and clinically relevant Clavien-Dindo complications (CR-CDC) were registered. Thirteen previously published risk prediction methods were included in the comparison, such as pancreatic duct diameter, palpable texture of pancreas, Braga score (BC), Fistula Risk Score, Modified Fistula Risk Score, Alternative Fistula Risk Score and multiple radiological parameters. ROC-curves were calculated to compare sensitivity and specificity for identifying high risk patients for CR-POPF and CR-CDC. The three most accurate risk prediction methods for CR-POPF were ACM (sensitivity 88.9%, specificity 52.6%; p=0.043), BC (87.5%, 56.6%; p=0.039) and visceral fat area to subcutaneous fat area ratio (75.5%, 80.0%; p=0.032). In predicting CR-CDC the three most accurate methods were ACM (73.9%, 56.2%; p=0.033), BC (68.4%, 59.5%; p=0.036) and TPAI (78.3%, 41.7%; p=0.012). ACM was shown to be as good as the more complicated risk scoring methods in the prediction of CR-POPF. It was good also in predicting all clinically relevant complications. ACM is easy to use during operation and can be recommended as a routine risk prediction method.

Comparison of prognostic models for burn patients: A retrospective nationwide registry study

BackgroundPrognostic burn index (PBI) is a unique model utilized to predict mortality of burn patients in Japan. In contrast, other prediction models are rarely used in Japan, and their accuracy and predictive value are unknown. The present study aimed to compare commonly used burn prediction models and determine the appropriate model for mortality prediction in Japanese burn patients. MethodsJapanese burn patients registered in the nationwide burn registry of Japanese Society for Burn Injury between April 1, 2011 and March 31, 2019 were reviewed retrospectively. The prognostic performance of PBI was compared with Baux score, revised Baux score, abbreviated burn severity index (ABSI), Ryan score and Belgian outcome in burn injury score (BOBI). The primary outcome was in-hospital mortality. ResultsThe study included 7911 acute burn patients. The overall mortality rate was 10.7%, the median age was 52 (interquartile range, 26–72) years, and the median % total body surface area was 7% (interquartile range, 3%–17%). The areas under the receiver operating characteristic curve for PBI, Baux score, ABSI, revised Baux score, Ryan score, and BOBI were 0.940 (95% confidence interval [CI]: 0.931–0.948), 0.943 (95% CI: 0.934–0.951; p=0.002), 0.945 (95% CI: 0.937–0.953; p=0.058), 0.946 (95% CI: 0.937–0.953; p=0.002), 0.859 (95% CI: 0.846–0.870; p<0.001), and 0.896 (95% CI: 0.885–0.905; p<0.001), respectively. ConclusionAlthough the performance of PBI was good, it was not superior to the Baux score, revised Baux score, and ABSI. These three scores have a high prognostic accuracy. Hence, they are considered as alternative burn prognostic scores in Japan. The Baux score was an optimal prognostic model for patients with burns in Japan.

Proposal for a clinicopathological prognostic score for resected gastric cancer patients.

Background:Factors other than pTNM stage have been associated with gastric cancer (GC) prognosis, and several alternative prognostic scores have been constructed. Our aims are to identify prognostic factors in western GC patients and to build clinicopathological prognostic models for overall survival (OS) and disease-free survival (DFS).Methods:A Retrospective study of 204 cases of GC resected during the years 2000 to 2014 was conducted in our hospital. Clinicopathological features were assessed, univariate and multivariate analysis were performed and prognostic scores were constructed.Results:Most patients were diagnosed at pTNM stages II and III (36.9% and 48.1%, respectively). According to Laurén classification, tumors were intestinal (55.8%), diffuse (35.2%) and mixed (9%). During follow-up, 43.5% of patients had tumor recurrence, and 28.6% died due to tumor. Univariate analysis showed that patient age, Laurén subtype, signet-ring cell morphology, pTNM stage, tumor grade, perineural invasion, growth pattern, intratumoral inflammation, adjuvant therapy, and desmoplasia were significantly related to tumor progression or death. Multivariate analysis showed that Laurén subtype, pT stage, and lymph node ratio (LNR) were significantly and independently associated with GC recurrence. Laurén subtype and LNR were significantly related to patient survival. Prognostic scores for tumor progression and death were developed and patients were classified into four prognostic groups which showed good prognostic performance.Conclusion:A prognostic model comprising histological features such as Laurén subtype can be easily applied in clinical practice, and provides more prognostic information than pTNM stage alone. These models can further stratify resected GC patients and have the potential to aid in the individualization of patient management.

Decision support for the Ecosystem-Based Management of a Range-Extending Species in a Global Marine Hotspot Presents Effective Strategies and Challenges

Climate-driven changes in ocean currents have facilitated the range extension of the long-spined sea urchin (Centrostephanus rodgersii) from Australia’s mainland to eastern Tasmania over recent decades. Since its arrival, the destructive grazing of the urchin has led to widespread formation of sea urchin ‘barrens’. The loss of habitat, biodiversity and productivity for important commercial reef species in conjunction with the development of an urchin fishery has led to conflicting objectives among some stakeholders that pose complex challenges for regional management. Stakeholders representatives and managers were engaged via a participatory workshop and subsequent one-on-one surveys to trial a structured decision making process to identify effective ecosystem-based management strategies. We directly and indirectly elicited each preferences for nine alternative management strategies by presenting them with the 10-year consequences of each strategy estimated from an ecosystem model of Tasmanian reef communities. These preferences were included in cost-effectiveness scores that were averaged (across stakeholders) to enable strategy ranking from most to least cost-effective. Rankings revealed strategies that included sea urchin removal or translocation of predatory lobsters were the most cost-effective. However, assessment of stakeholders’ individual cost-effectiveness scores showed some disparity among preferences in high ranking strategies. Additionally, inconsistencies in strategy preferences using alternative (direct or indirect) ranking scores reveal conflicting objectives as the most plausible explanation. Our study illustrates how structured decision making can effectively facilitate ecosystem-based management by engaging stakeholders step-by-step towards management strategy implementation and promoting collective learning.

External validation of International Prognostic Score for asymptomatic early stage chronic lymphocytic leukaemia and proposal of an alternative score.

Most patients with chronic lymphocytic leukaemia (CLL) are nowadays diagnosed without any symptoms and do not require therapy. A prognostic score identifying patients within this large group who are at high risk of disease progression would be highly beneficial. The recently published International Prognostic Score for Early asymptomatic patients (IPS-E) uses combination of absolute lymphocyte count (ALC) >15×109 /l, palpable lymphadenopathy, and unmutated immunoglobulin heavy-chain variable-region (IGHV) gene to predict the time to first-line therapy (TTFT). Patients at low, intermediate, and high risk had estimated 5-year TTFT of 8%, 28%, and 61%. We performed an external validation of the IPS-E score using an unselected, consecutive group of 130 Binet A patients. The 5-year TTFT was 11%, 36%, and 78% (C-statistic 0·74). Furthermore, we propose an alternative system (AIPS-E) using cytogenetic aberrations instead of palpable lymphadenopathy. This system yielded 5-year TTFT of 14%, 40%, and 72%. These results were externally validated in 388 Binet A patients from five Czech centres; the 5-year TTFT was 16%, 37%, and 80% (C-statistic 0·74). In conclusion, we have successfully validated the IPS-E score for patients with early stage CLL. In addition, we propose a modified scoring system, the AIPS-E, combining IGHV, fluorescence in situ hybridisation, and ALC.