scholarly journals Machine Learning Prediction of Progression in FEV1 in the COPDGene Study

Author(s):  
Adel Boueiz ◽  
Zhonghui Xu ◽  
Yale Chang ◽  
Aria Masoomi ◽  
Andrew Gregory ◽  
...  

Background: The heterogeneous nature of COPD complicates the identification of the predictors of disease progression and consequently the development of effective therapies. We aimed to improve the prediction of disease progression in COPD by using machine learning and incorporating a rich dataset of phenotypic features. Methods: We included 4,496 smokers with available data from their enrollment and 5-year follow-up visits in the Genetic Epidemiology of COPD (COPDGene) study. We constructed supervised random forest models to predict 5-year progression in FEV1 from 46 baseline demographic, clinical, physiologic, and imaging features. Using cross-validation, we randomly partitioned participants into training and testing samples. We also validated the results in the COPDGene 10-year follow-up visit. Results: Predicting the change in FEV1 over time is more challenging than simply predicting the future absolute FEV1 level. Nevertheless, the area under the ROC curves for the prediction of subjects in the top quartile of observed disease progression was 0.70 in the 10-year follow-up data. The model performance accuracy was best for GOLD1-2 subjects and it was harder to achieve accurate prediction in advanced stages of the disease. Predictive variables differed in their relative importance as well as for the predictions by GOLD grade. Conclusion: This state-of-the art approach along with deep phenotyping predicts FEV1 progression with reasonable accuracy. There is significant room for improvement in future models. This prediction model facilitates the identification of smokers at increased risk for rapid disease progression. Such findings may be useful in the selection of patient populations for targeted clinical trials.

RMD Open ◽  
2021 ◽  
Vol 7 (2) ◽  
pp. e001524
Author(s):  
Nina Marijn van Leeuwen ◽  
Marc Maurits ◽  
Sophie Liem ◽  
Jacopo Ciaffi ◽  
Nina Ajmone Marsan ◽  
...  

ObjectivesTo develop a prediction model to guide annual assessment of systemic sclerosis (SSc) patients tailored in accordance to disease activity.MethodsA machine learning approach was used to develop a model that can identify patients without disease progression. SSc patients included in the prospective Leiden SSc cohort and fulfilling the ACR/EULAR 2013 criteria were included. Disease progression was defined as progression in ≥1 organ system, and/or start of immunosuppression or death. Using elastic-net-regularisation, and including 90 independent clinical variables (100% complete), we trained the model on 75% and validated it on 25% of the patients, optimising on negative predictive value (NPV) to minimise the likelihood of missing progression. Probability cutoffs were identified for low and high risk for disease progression by expert assessment.ResultsOf the 492 SSc patients (follow-up range: 2–10 years), disease progression during follow-up was observed in 52% (median time 4.9 years). Performance of the model in the test set showed an AUC-ROC of 0.66. Probability score cutoffs were defined: low risk for disease progression (<0.197, NPV:1.0; 29% of patients), intermediate risk (0.197–0.223, NPV:0.82; 27%) and high risk (>0.223, NPV:0.78; 44%). The relevant variables for the model were: previous use of cyclophosphamide or corticosteroids, start with immunosuppressive drugs, previous gastrointestinal progression, previous cardiovascular event, pulmonary arterial hypertension, modified Rodnan Skin Score, creatine kinase and diffusing capacity for carbon monoxide.ConclusionOur machine-learning-assisted model for progression enabled us to classify 29% of SSc patients as ‘low risk’. In this group, annual assessment programmes could be less extensive than indicated by international guidelines.


2019 ◽  
Vol 74 (12) ◽  
pp. 1903-1909 ◽  
Author(s):  
Meredith L Wallace ◽  
Daniel J Buysse ◽  
Susan Redline ◽  
Katie L Stone ◽  
Kristine Ensrud ◽  
...  

Abstract Background Sleep characteristics related to duration, timing, continuity, and sleepiness are associated with mortality in older adults, but rarely considered in health recommendations. We applied machine learning to: (i) establish the predictive ability of a multidimensional self-reported sleep domain for all-cause and cardiovascular mortality in older adults relative to other established risk factors and (ii) to identify which sleep characteristics are most predictive. Methods The analytic sample includes N = 8,668 older adults (54% female) aged 65–99 years with self-reported sleep characterization and longitudinal follow-up (≤15.5 years), aggregated from three epidemiological cohorts. We used variable importance (VIMP) metrics from a random survival forest to rank the predictive abilities of 47 measures and domains to which they belong. VIMPs > 0 indicate predictive variables/domains. Results Multidimensional sleep was a significant predictor of all-cause (VIMP [99.9% confidence interval {CI}] = 0.94 [0.60, 1.29]) and cardiovascular (1.98 [1.31, 2.64]) mortality. For all-cause mortality, it ranked below that of the sociodemographic (3.94 [3.02, 4.87]), physical health (3.79 [3.01, 4.57]), and medication (1.33 [0.94, 1.73]) domains but above that of the health behaviors domain (0.22 [0.06, 0.38]). The domains were ranked similarly for cardiovascular mortality. The most predictive individual sleep characteristics across outcomes were time in bed, hours spent napping, and wake-up time. Conclusion Multidimensional sleep is an important predictor of mortality that should be considered among other more routinely used predictors. Future research should develop tools for measuring multidimensional sleep—especially those incorporating time in bed, napping, and timing—and test mechanistic pathways through which these characteristics relate to mortality.


Rheumatology ◽  
2020 ◽  
Vol 59 (11) ◽  
pp. 3390-3399
Author(s):  
Alan M Rathbun ◽  
Michelle D Shardell ◽  
Alice S Ryan ◽  
Michelle S Yau ◽  
Joseph J Gallo ◽  
...  

Abstract Objectives Osteoarthritis (OA) disease progression may lead to deteriorating psychosocial function, but it is unclear what aspects of disease severity are related to the onset of depression. This study assessed which components of OA disease progression cumulatively contribute to depression onset in persons with radiographic knee OA. Methods Osteoarthritis Initiative participants (n = 1651) with radiographic disease (Kellgren-Lawrence grade ≥2) in one or both knees and below the screening threshold for probable depression [Center for Epidemiological Studies Depression (CES-D) scale &lt;16] at baseline were included. Disease severity was measured from baseline to the third annual follow-up visit using joint space width, 20-meter gait speed, and the Western Ontario and McMaster Universities Osteoarthritis Index pain subscale, each categorized into quintiles. Depression onset (CES-D ≥ 16) was assessed annually at four follow-up visits. Marginal structural models that account for time-dependent confounding and attrition evaluated the association between each time-varying disease severity measure and depression onset. Results Each disease severity measure exhibited a non-linear relationship concerning the probability of depression onset, with the higher quintiles generally being associated with a larger risk. The highest quintile (relative to the lowest) of joint space width and gait speed were both significantly associated with depression onset. By contrast, none of the higher pain quintiles compared with the lowest were significantly associated with the onset of depression. Conclusion Faster disease progression as measured by either worsening structural severity or decreasing physical performance corresponds to an increased risk of depression among individuals with radiographic knee OA.


Blood ◽  
2012 ◽  
Vol 120 (21) ◽  
pp. 5060-5060
Author(s):  
Grace Kam ◽  
Richard Yiu ◽  
Ai Leen Ang ◽  
Yvonne SM Loh ◽  
Yeh Ching Linn ◽  
...  

Abstract Abstract 5060 Less than 20% of patients with essential thrombocythemia (ET) are diagnosed below the age of 60. Patients with ET have increased risk of thrombosis and bleeding and potential for progression to myelofibrosis (MF) or acute myeloid leukaemia (AML). In limited studies of young patients, the clinical course has been relatively benign with low rates of transformation to AML or MF. Thrombohemorrhagic events are generally few, but higher than that of the general population. This study aims to characterize of a group ET patients diagnosed at age ≤40, their thrombotic and hemorrhagic events, disease progression and treatment given. Patients were identified through a single institution MPN registry. This is an IRB approved registry that captures comprehensive information about patients with ET. Data on patient demographics, treatment, and disease-related events were obtained. Patients were diagnosed from 1975–2011, using either WHO or PVSG criteria depending on date of diagnosis. Kaplan-Meier method was used for survival analysis. 59 patients were diagnosed with ET at age ≤40. Median age of diagnosis was 31. 5years (range 16–40), with a median follow up of 7. 7years (0. 4–33. 8). All were of Asian descent: 81. 4% Chinese, 11. 9% Malay, 3. 4% Indian and 3. 4% Filipino. 40. 7% were male. JAK2 V617F mutation was screened for in 61%. Of these patients, 11 were positive, 25 negative for the mutation. Mean presenting counts were: WBC 10. 7 × 109/L (5. 9–21. 3), Hb 13. 6g/dL (9. 7–16. 4), platelets 957 × 109/L (449–2377). Splenomegaly was noted in 3 patients. 20. 3% had underlying hypertension, 16. 9% hyperlipidemia and 5. 1% diabetes mellitus. One patient had a prior stroke. Another had prior portal vein thrombosis. At diagnosis, 23. 7% were symptomatic, with microvascular symptoms of headache (11. 9%) and giddiness (6. 8%) being most common. The remainder were diagnosed incidentally, on health screening or when seeking medical attention for unrelated conditions. One patient presented with a myocardial infarction at diagnosis, while another had a significant bleeding post hemorrhoidectomy with drop in Hb by >2g/dL (platelet 2457 × 109/L). Based on a history of prior thrombosis, 3 patients were defined as high risk for thrombotic events. 67. 8% of patients had cytoreduction, indications being platelets ≥1500 × 109/L (n=16), presence of risk factors for atherosclerotic disease (n=11) and history/onset of thrombosis (n=5). In 8, the reason for cytoreduction could not be ascertained. Hydroxyurea was most commonly used (62. 7%), followed by anagrelide in 52. 5% and interferon 25. 4%. 5. 1% received busulphan, and 1. 7% 32P. Use of antiplatelet therapy was noted in 83. 8%, most frequently aspirin (76. 5%) and ticlopidine (11. 9%). On follow up, 2 arterial thromboses occurred (stroke, TIA), giving a thrombosis rate of 0. 39%/patients/year. Neither was a recurrent thrombosis. No venous thrombosis or major bleeds occurred. 20. 4% had minor mucocutaneous bleeding; 5 had platelets ≥1500 × 109/L at that time. 3. 4% had disease progression due to MF and another 3. 4% had AML. 3. 4% of patients died due to AML. Median survival was 33. 8years (95% confidence interval 30. 3–35. 5). Initial blood counts, presence of JAK2 and high risk disease status did not correlate with thrombotic risk, risk of death or disease progression. Use of antiplatelet agents and a platelet count ≥1500 × 109/L did not correlate with bleeding risk. Few studies have looked exclusively at young patients with ET. In this group, most patients were asymptomatic and well, ET being diagnosed incidentally. They were predominantly at low risk for thrombosis and other ET-related complications. The period of follow up was comparable to that of other studies and during that time, the rate of complications and risk of disease progression was low. The thrombosis rate of 0. 39% per patient year was less than that reported by other groups (2. 2–2. 6 thromboses/100patients/year) (Leukaemia 2007;21:1218–1223, Clin Appl Thrombosis/Hemostasis 2000;6(1):31–35) but similar to the 0. 74%/patient year reported by Barbui (Blood. Epub. June 13 2012). Overall findings generally complemented those reported by other groups. No risk factors were found to influence the occurrence of complications, but the number of events was small. Follow up of this group of patients over time is essential to see if their disease course remains benign or if complications will increase with time. Soli Deo Gloria Disclosures: Kam: Shire Pharmaceuticals: Consultancy, grant to support the MPN registry Other.


2020 ◽  
Vol 17 (4) ◽  
Author(s):  
Nan Yu ◽  
Yong Yu ◽  
Shubo Cai ◽  
Cong Shen ◽  
Youmin Guo

Objectives: To describe the characteristics of computed tomography (CT) in patients with 2019 novel coronavirus (COVID-19) pneumonia and their changes during disease progression. Patients and Methods: A total of 96 chest CT scans of 61 pneumonia patients associated with COVID-19 were reviewed to identify CT features associated with the time of symptom onset and the evolution of disease. Results: The initial CTs of 61 patients were obtained during 1 to 11 days after the onset. The main CT pattern of initial CT obtained during 1 - 3 days after the symptom onset was single (7/23, 35%) or multiple ground-glass opacity (GGO, 8/23, 35%). At 4 - 7 days after the symptom onset, the main imaging features were crazy paving GGO mixed with partial consolidation pattern (15/32, 47%). At 8 - 11 days after the symptom onset, the CT images showed consolidation pattern (3/6, 50%). A total of 35 follow up CTs were collected. The mean interval time between each follow up CT was 3 ± 2 days. The CT patterns also changed with the evolution of the disease: the features of GGO manifested at the early stage (1 - 3d). The crazy paving GGO pattern, consolidation pattern and mixed with partial consolidation pattern were found 4 to 14 days after the onset. In the absorption stage (15 - 24d), both density and extent of lesions were reduced. Conclusion: The CT imaging features are associated with the time of symptom onset and evolution of disease. Understanding the imaging characteristics of each stage is very helpful for understanding the development of disease.


2013 ◽  
Vol 31 (15_suppl) ◽  
pp. e16066-e16066
Author(s):  
Rohini Khorana Hernandez ◽  
Johan Mesterton ◽  
Jonas Banefelt ◽  
Jan Stålhammar ◽  
Patrik Sobocki ◽  
...  

e16066 Background: ADT is the standard of care in Sweden for PC pts with signs of recurrence after primary therapy (tx). Studies of predictors of metastasis (mets) and survival have largely focused on pt characteristics at cancer diagnosis. Time-varying factors, such as prostate-specific antigen (PSA) levels, may have greater impact on a pt’s risk of disease progression. This study examines predictors of mets and survival among men with PC treated with ADT. Methods: Using electronic medical records from Swedish urology clinics linked to national registries (Cancer Registry, National Pt Registry, Cause of Death Registry), we identified men with PC and no evidence of mets treated with ≥6 months (mos) ADT (gonadotropin-releasing hormone agonists/antagonists or bilateral orchiectomy) between 2000-2010 with ≥2 PSA values. Men were followed from ADT index date to mets, death, or end of follow-up (12/31/2010). Multivariate competing risks regression analysis was used to estimate hazard ratios (HR) and 95% CIs; predictors and covariates of interest included PC diagnosis year (yr), age, comorbidities, anti-androgen tx, region, and time-varying characteristics (PSA absolute value, PSA doubling time [DT]). Results: Cohort was 446 men with mean follow-up of 3.3 yrs. Most mets were to the bone (7-yr cumulative incidence 25% for bone, 30% for any mets). Median survival was 6 yrs (5.9 mos after bone mets, 6.1 mos after any mets). Higher PSA and shorter PSA DT were strong predictors of all outcomes. In particular, PSA DT ≤ 6 mos was associated with increased risk of bone mets (13.9 [8.0 – 24.1]), any mets (7.9 [4.9 – 12.8]), mortality (5.7 [3.9 – 8.5]), and bone mets-free survival (6.9 [4.7–10.1]) when compared to PSA DT > 6 mos. HRs were adjusted for age, Charlson comorbidity index, anti-androgen tx, and region. Conclusions: PC pts treated with ADT are at significant risk of bone mets, any mets, and death. This study based on real-world data demonstrates the importance of PSA measured after ADT initiation in defining high risk of these outcomes, particularly PSA DT ≤6 mos. PC pts may benefit from new tx to prevent disease progression since survival is short after bone or other mets.


Stroke ◽  
2013 ◽  
Vol 44 (suppl_1) ◽  
Author(s):  
Shyian S Jen ◽  
Elizabeth M Sweeney ◽  
Argye E Hillis ◽  
Ciprian M Crainiceanu ◽  
John W Krakauer ◽  
...  

Background: Recent studies have detected a population of acute ischemic stroke patients whose MRI profile is associated with intracranial hemorrhage (ICH) when treated beyond 3 hours. This so-called malignant profile (MP) supports MRI based selection of patients for treatment. Purpose: To test the hypothesis that there is an MRI based volumetric profile that identifies patients at increased risk of ICH when treated with IV tPA within 3 hours Methods: An analysis was performed on a database of stroke patients provided by the STIR and VISTA Imaging Investigators. 75 patients were identified who had DWI, PWI, and GRE images prior to IV tPA and follow-up imaging to assess for parenchymal hematoma (PH). The pre-tPA MRI scans were analyzed using Matlab software to calculate DWI and PWI volumes. DWI lesions were defined by an ADC threshold of 600. PWI lesions were defined by a time-to-peak threshold of an 8 seconds delay. Follow-up GRE images were reviewed for evidence of PH. ROC curves were generated using thresholds from 1-300mL. Results: 44 of the 75 patients were women with a mean ± stdev age of 70±17. The mean NIHSS was 12±9. The mean time from stroke onset to tPA administration was 147±30 minutes. For the entire cohort, mean lesion volumes were 22±41mL for DWI and 41±42 mL for PWI. 9 patients developed PH. For the PH group, mean lesion volumes were 24±20 mL for DWI and 48±40 mL for PWI. The ROC curves are shown in Figure 1. The areas under the curves are 0.68 for DWI volumes and 0.60 for PWI volumes. The optimal thresholds for predicting PH were 13 mL for DWI with a sensitivity of 0.67 and a specificity of 0.66 and 21 mL for PWI with a sensitivity of 0.78 and a specificity of 0.46. Conclusions: Although DWI volume performed better than PWI volume in predicting PH, neither served as a robust predictor in this population. Although further studies are needed, these results suggest that an MRI profile defined by DWI and PWI volumes in the 0-3 hour window may not be able to reliably guide clinical management.


Author(s):  
Douglas L Zentner ◽  
Joshua K Raabe ◽  
Timothy K Cross ◽  
Peter C Jacobson

Scale and hierarchy have received less attention in aquatic systems compared to terrestrial. Walleye Sander vitreus spawning habitat offers an opportunity to investigate scale’s importance. We estimated lake-, transect-, and quadrat-scale influences on nearshore walleye egg deposition in 28 Minnesota lakes from 2016-2018. Random forest models (RFM) estimated importance of predictive variables to walleye egg deposition. Predictive accuracies of a multi-scale classification tree (CT) and a quadrat-scale CT were compared. RFM results suggested that five of our variables were unimportant when predicting egg deposition. The multi-scale CT was more accurate than the quadrat-scale CT when predicting egg deposition. Both model results suggest that in-lake egg deposition by walleye is regulated by hierarchical abiotic processes and that silt/clay abundance at the transect-scale (reef-scale) is more important than abundance at the quadrat-scale (within-reef). Our results show machine learning can be used for scale-optimization and potentially to determine cross-scale interactions. Further incorporation of scale and hierarchy into studies of aquatic systems will increase our understanding of species-habitat relationships, especially in lentic systems where multi-scale approaches are rarely used.


Author(s):  
Rui Fu ◽  
Jiamin Shi ◽  
Michael Chaiton ◽  
Adam M Leventhal ◽  
Jennifer B Unger ◽  
...  

Abstract Introduction Machine learning presents a unique opportunity to improve electronic cigarette (vaping) monitoring in youth. Here we built a random forest model to predict frequent vaping status among Californian youth and to identify contributing factors and vulnerable populations. Methods In this prospective cohort study, 1,281 ever-vaping twelfth-grade students from metropolitan Los Angeles were surveyed in Fall and in 6-month in Spring. Frequent vaping was measured at the 6-month follow-up as nicotine-containing vaping on 20 or more days in past 30 days. Predictors (n=131) encompassed sociodemographic characteristics, substance use and perceptions, health status, and characteristics of the household, school and neighborhood. A random forest was developed to identify the top ten predictors of frequent vaping and interactions by sociodemographic variables. Results Forty participants (3.1%) reported frequent vaping at the follow-up. The random forest outperformed a logistic regression model in prediction (C-Index=0.87 vs. 0.77). Higher past-month nicotine concentration in vape, more daily vaping sessions, and greater nicotine dependence were the top three of the ten most important predictors of frequent vaping. Interactions were found between age and perceived discrimination, and between age and race/ethnicity, as those who were younger than their classmates and either reported experiencing discrimination frequently or identified as Asian or Native American/Pacific Islander were at increased risk of becoming frequent vapers. Conclusions Machine learning can produce models that accurately predict progression of vaping behaviours among youth. The potential association between frequent vaping and perceived discrimination warrants more in-depth analyses to confirm if discrimination constitutes a cause of increased vaping.


Blood ◽  
2012 ◽  
Vol 120 (21) ◽  
pp. 2147-2147
Author(s):  
Caroline Besson ◽  
Sophie Prevot ◽  
Houria Chavez ◽  
Selma Trabelsi ◽  
Michele Genin ◽  
...  

Abstract Abstract 2147 Background: Human Immunodeficiency Virus (HIV) infection is associated with an increased risk of Hodgkin lymphoma (HL) and B-cell non-Hodgkin lymphoma (NHL). Increased risk of NHL is strongly correlated to the severity of the underlying immunodeficiency. Introduction of combined antiretroviral therapy (cART) has reduced the incidence of NHL -but not of HL's- among HIV-infected individuals. Outcomes are reported to be poorer among HIV-infected patients with HL or NHL than among non-HIV-infected patients. We carry out a cohort with the aim to study the characteristics and outcome of HIV-related lymphomas. Methods: The multicentric prospective Cohort of HIV related lymphomas (ANRS-CO16 Lymphovir cohort) enrolled 116 adult patients in 32 centers between October 2007 and April 2012. Investigations were performed after approval of the ethic committee. Patients were included at diagnosis of lymphoma (n=108) or at first relapse (1 HL, 7 NHL). Data collection concerned HIV infection history, clinical, biological and histological presentation, treatment and evolution of lymphoma. Pathological materials were centralized and 91 cases were reviewed. Diagnoses were based on World Health Organization criteria. Each patient was followed every 6 months during 5 years. Results: Among the 116 patients, 39.7% (46) were diagnosed with HL and 60.3% (70) with NHL. Median age was 43.5 years (ranging from 20 to 61) among patients with HL and 47 years (23 to 67) among those with NHL. Gender (male/female) ratio was 8.2 (41/5) among patients with HL, 1.7 (44/26) among those with NHL. The histological distribution of NHL were diffuse large B-cell lymphoma (DLBCL) 54.3% (38), Burkitt lymphomas (BL) 18.6% (13), plasmablastic lymphoma 10% (7), marginal zone/lymphoplasmocytic lymphoma 7.1% (5), others 10%: PTLD- like lymphoma (2), primary effusion lymphoma (1), follicular lymphoma (1), anaplastic lymphoma (1), unclassified (2). There was a predominance of clinical stages III/IV versus I/II among HL (76.7%, 33/43) and NHL patients (73.5%, 50/68). Among patients with DLBCL, LDH level was elevated in 68.4% (26/38) and performance status altered (2–4 versus 0–1) in 38.5% (15/39). HIV infection had been diagnosed for a median of 151 months (0 to 312) among HL patients and 117 (0 to 327) among those with NHL. The interval between diagnoses of HIV infection and lymphoma was shorter than 3 months for 2 patients with HL and 13 with NHL. All other patients except 6 NHL patients had been treated with ART at diagnosis of lymphoma. Median durations of ARV were 128 months (2 to 238) among HL patients and 119 months (1 to 236) among those with NHL. Patients with HL had a median CD4 T-cell count at diagnosis of lymphoma of 353/mm3 (range 37–1120), those with NHL, 261/mm3 (range 7–1322)]. The median interval between lymphoma occurrence and last follow-up was 21 months (range 0–41). During follow-up, all patients were treated with ARV. Among first-line HL patients, 39 out of 40 were treated with ABVD. Out of 40 patients with DLBCL or BL, 30 received chemotherapies combined with rituximab. At 24 months, overall survival is 95% among patients with HL and 78% among those with NHL (Figure 1). Two HL patients died during follow-up: one HL patient included in relapse from progression, another from a second cancer. Sixteen NHL patients died during follow-up: there were 10 early deaths (<6 months) from complications of treatment (9) or disease progression (1) and 6 later deaths from disease progression (4), second cancer (1), unknown (1). None of the patients who died during the first 6 months following diagnosis had received rituximab. Conclusions: The present study points out the high proportion of HL among HIV infection with lymphoma in the cART era and their favourable outcome compared to previous reports. This study also strengthens the heterogeneity of HIV-related lymphomas and the frequency of early deaths among patients with NHL. Disclosures: No relevant conflicts of interest to declare.


Sign in / Sign up

Export Citation Format

Share Document