FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W1972774340> ?p ?o ?g. }

• W1972774340 endingPage "1515" @default.
• W1972774340 startingPage "1512" @default.
• W1972774340 abstract "Etiological research aims to investigate the causal relationship between putative risk factors (or determinants) and a given disease or other outcome. In contrast, prognostic research aims to predict the probability of a given clinical outcome and in this perspective the pathophysiology of the disease is not an issue. Multivariate modeling is a fundamental tool both to infer causality and to investigate prognostic factors in epidemiological research. The analytical approaches to etiological and prognostic studies are strictly dependent on the research question and imply knowledge of the main statistical procedures for model building and data interpretation. In this paper we describe the application of multivariate statistical modeling in etiological and prognostic research. We will mainly focus on the differences in model building and data interpretation between these two areas of epidemiologic research. Etiological research aims to investigate the causal relationship between putative risk factors (or determinants) and a given disease or other outcome. In contrast, prognostic research aims to predict the probability of a given clinical outcome and in this perspective the pathophysiology of the disease is not an issue. Multivariate modeling is a fundamental tool both to infer causality and to investigate prognostic factors in epidemiological research. The analytical approaches to etiological and prognostic studies are strictly dependent on the research question and imply knowledge of the main statistical procedures for model building and data interpretation. In this paper we describe the application of multivariate statistical modeling in etiological and prognostic research. We will mainly focus on the differences in model building and data interpretation between these two areas of epidemiologic research. In two previous papers of this series we described linear, logistic, and Cox regression analyses1.Tripepi G. Jager K.J. Dekker F.W. et al.Linear and logistic regression analysis.Kidney Int. 2008; 73: 806-810Abstract Full Text Full Text PDF PubMed Scopus (65) Google Scholar,2.van Dijk P.C. Jager K.J. Zwinderman A.H. et al.The analysis of survival data in nephrology. Basic concepts and methods of Cox regression.Kidney Int. 2 July 2008Abstract Full Text Full Text PDF Scopus (59) Google Scholar as fundamental tools for investigating the relationship between exposure to risk factors and clinical outcomes. According to the research question being addressed, statistical models can be used to test both etiological (that is to infer causality) and prognostic hypotheses (that is to predict a given clinical outcome). In etiological research we are interested in determining the presence or absence of a presumed causal relationship between a putative risk factor and a specific clinical condition and in this particular perspective confounding and previous knowledge of the pathophysiology of the disease play a fundamental role. In descriptive prognostic research we aim at predicting the risk (that is the probability) of disease without any concern about causality and confounding. Multivariate modeling in etiological research may serve two scopes: to adjust for confounders when confounders are variables not in the same causal pathway leading to a given outcome and to adjust for variables within a certain causal pathway to unravel mechanisms (see below). Framingham study investigators were the first to apply statistical modeling to the study of complex (multifactorial) diseases like hypertension or atherosclerosis. What matters in etiological research is establishing the causal involvement of one or more risk factors in a given disease. Thus, when we are to test the association of a putative risk factor, for example obesity, as a cause of coronary heart disease, it is fundamental that we establish whether the association between this factor and the outcome (for example incident myocardial infarction) is independent of other confounding risk factors. To this end we construct statistical models including all risk factors whose causal role in determining the outcome is reasonably well demonstrated. Otherwise stated, we set up an experiment where we control by multivariate analysis all possible sources of confounding.3.Jager K.J. Zoccali C. Macleod A. Dekker F.W. Confounding: what it is and how to deal with it.Kidney Int. 2008; 73: 256-260Abstract Full Text Full Text PDF PubMed Scopus (246) Google Scholar In this perspective, in the process of selecting potential confounders, particular attention is needed to make sure that the variable in question, rather than a confounder, represents a factor in the same chain of events leading to the outcome. For instance, insulin resistance is a well-recognized effect of overweight/obesity. If we adjust the analysis for insulin resistance we may reduce or even cancel out the link between obesity and myocardial infarction. Therefore, we should not include measures of insulin sensitivity into the multivariate model testing obesity as a causal factor for myocardial infarction. On the other hand, if we are interested to understand mechanisms, that is to establish whether obesity induced by mechanisms other than insulin resistance determines myocardial infarction, then adjustment for insulin resistance may assist to explore this hypothesis. Similarly in an animal study testing whether obesity may cause cardiovascular (CV) damage by mechanisms other than insulin resistance, three interventions can be compared: a control intervention (that is a normal calorie diet), a high calorie diet (that makes normal animals obese and insulin resistant), and the same diet administered to genetically modified animals which maintain normal insulin sensitivity while developing obesity. Of course the three groups are matched for relevant baseline variables such as age and sex. In this experiment the investigator is ideally poised for isolating the effect of obesity and insulin resistance on the CV system because all factors, but the interventions, are equalized among experimental groups. Similarly, in epidemiologic research we isolate the effect of the exposure/non exposure to the risk factor in question (obesity vs normal weight, insulin resistance vs normal insulin sensitivity) by adjusting for the confounding effect of other risk factors (age, sex, smoking, and others). In the nineties, it was hypothesized that subclinical Chlamydia pneumoniae infection is a risk factor for CV disease. Observations in chronic ambulatory peritoneal dialysis (CAPD) patients specifically implicated this pathogen in the high risk of death of these patients.4.Haubitz M. Brunkhorst R. C-reactive protein and chronic Chlamydia pneumoniae infection-long-term predictors for cardiovascular disease and survival in patients on peritoneal dialysis.Nephrol Dial Transplant. 2001; 16: 809-815Crossref PubMed Scopus (96) Google Scholar As this study was relatively small and could not adequately control for confounding factors, the problem was re-examined in a cohort study in 227 end-stage renal disease (ESRD) patients.5.Zoccali C. Mallamaci F. Tripepi G. et al.Chlamydia pneumoniae, overall and cardiovascular.Kidney Int. 2003; 64: 579-584Abstract Full Text Full Text PDF PubMed Scopus (25) Google Scholar To analyze the results of the study, the authors divided patients into three groups on the basis of immunoglobulin A (IgA) anti-Chlamydia antibody titer (from seronegativity to high levels) and investigated the etiological role of Chlamydia infection for all-cause and CV mortality by Cox regression analysis (Table 1).2.van Dijk P.C. Jager K.J. Zwinderman A.H. et al.The analysis of survival data in nephrology. Basic concepts and methods of Cox regression.Kidney Int. 2 July 2008Abstract Full Text Full Text PDF Scopus (59) Google ScholarTable 1Univariate Cox regression analysis of IgA anti-Chlamydia titer for all-cause and CV mortalityAll-cause mortalityCV mortalityHazard ratio, 95% CI and P-valueHazard ratio, 95% CI and P-valueIgA anti-Chlamydia titer Seronegative1aReference group.1aReference group. IgA titer 1:81.22 (0.64–2.34), P=0.551.65 (0.82–3.34), P=0.16 IgA titer ≥1:161.89 (1.25–2.87), P=0.0031.77 (1.05–2.98), P=0.03IgA, immunoglobulin A; CV, cardiovascular.a Reference group. Open table in a new tab IgA, immunoglobulin A; CV, cardiovascular. During the follow-up, 102 patients died, 68% of them of CV causes. On univariate Cox regression analysis (Table 1), there was an IgA anti-Chlamydia titer-dependent increase in the risk of all-cause and CV mortality so that patients with a high titer (≥1:16) displayed a significantly higher risk of all-cause (hazard ratio: 1.89; excess risk: +89%) and CV mortality (hazard ratio: 1.77, excess risk: +77%) when compared to seronegative patients. If Chlamydia infection is causally involved in the high risk of death in ESRD population the relationship between IgA anti-C. pneumoniae titer and all-cause and CV death should be unconfounded, that is independent of other risk factors. In this study, age and smoking appeared to be potential confounders3.Jager K.J. Zoccali C. Macleod A. Dekker F.W. Confounding: what it is and how to deal with it.Kidney Int. 2008; 73: 256-260Abstract Full Text Full Text PDF PubMed Scopus (246) Google Scholar for the interpretation of the link between Chlamydia and clinical outcomes because they were related to both IgA anti-Chlamydia titer (the exposure) and to all-cause and CV mortality (the outcomes). The independent association between Chlamydia infection and all-cause and CV mortality was therefore tested in Cox's models adjusting for age and smoking (Table 2).Table 2Multiple Cox regression analysis of IgA anti-Chlamydia titer for all cause and CV mortalityAll-cause mortalityCV mortalityHazard ratio, 95% CI and P value (adjusted for age and sex)Hazard ratio, 95% CI and P value (adjusted for age and sex)IgA anti-Chlamydia titer Seronegative1aReference group.1aReference group. IgA titer 1:81.03 (0.54–2.00), P=0.921.40 (0.69–2.85), P=0.36 IgA titer ≥ 1:161.21 (0.79–1.87), P=0.391.16 (0.68–1.98), P=0.58 Age (1 year)1.05 (1.03–1.07), P<0.0011.05 (1.03–1.07), P<0.001 Smoking (yes/no)1.90 (1.27–2.85), P=0.0022.24 (1.38–3.65), P=0.001IgA, immunoglobulin A; CV, cardiovascular.a Reference group. Open table in a new tab IgA, immunoglobulin A; CV, cardiovascular. After adjustment for age and smoking the hazard ratio for all-cause and CV mortality became not significant and reduced by 36% (from 1.89 to 1.21) and by 34% (from 1.77 to 1.16), respectively indicating a relevant degree of confounding by these risk factors. Thus, age and smoking engendered ‘positive confounding’, that is they determined an overestimation of the risk associated with Chlamydia infection. In other words, the reason why ESRD patients exposed to high IgA anti-Chlamydia titer had an almost twofold increase in the risk of death when compared to seronegative patients was in a large part explained by the fact that these patients were older and more frequently smokers as compared to seronegative patients. The relationship between an endogenous inhibitor of nitric-oxide synthase, asymmetrical dimethylarginine (ADMA), and the risk of all-cause mortality and fatal and non-fatal CV events was investigated in a cohort of 225 hemodialysis patients followed-up for an average time of 33±15 months.6.Zoccali C. Bode-Böger S. Mallamaci F. et al.Plasma concentration of asymmetrical dimethylarginine and mortality in patients with end-stage renal disease: a prospective study.Lancet. 2001; 358: 2113-2117Abstract Full Text Full Text PDF PubMed Scopus (966) Google Scholar The study population was divided into three groups according to 50th and 75th percentile of plasma ADMA and the relationship between ADMA and study outcomes was preliminarily investigated by univariate (unadjusted) Cox regression analysis (Table 3).Table 3Univariate Cox regression analysis of plasma ADMA for all-cause mortality and fatal and non-fatal CV eventsAll-cause mortalityFatal and non-fatal cardiovascular eventsHazard ratio, 95% CI and P-valueHazard ratio, 95% CI and P-valuePlasma ADMA <50th percentile (<0.95 μmol/l)1aReference group.1aReference group. 50–75th percentile (0.95–1.53 μmol/l)1.39 (0.81–2.39), P=0.231.94 (1.13–3.31), P=0.02 >75th percentile (>1.53 μmol/l)2.43 (1.47–4.04), P<0.0012.41 (1.42–4.07), P=0.001ADMA, asymmetrical dimethylarginine.a Reference group. Open table in a new tab ADMA, asymmetrical dimethylarginine. Univariate Cox regression analysis (Table 3) showed that the overall risk of death and fatal and non-fatal CV events was progressively higher from the 50th percentile onwards. This biological gradient suggests that plasma ADMA is causally implicated in adverse clinical outcomes in ESRD patients. To confirm the study hypothesis, the authors identified a series of potential confounders and constructed Cox regression models of increasing complexity. Here we report only the results of a model adjusting for age and sex (Table 4).Table 4Multiple Cox regression analysis of plasma ADMA for all-cause mortality and fatal and non-fatal CV eventsAll-cause mortalityFatal and non fatal cardiovascular eventsHazard ratio, 95% CI and P-valueHazard ratio, 95% CI and P-valuePlasma ADMA <50th percentile (<0.95 μmol/l)1aReference group.1aReference group. 50–75th percentile (0.95–1.53 μmol/l)1.72 (1.00–2.97), P=0.052.13 (1.24–3.65), P=0.006 >75th percentile (>1.53 μmol/l)3.11 (1.83–5.27), P<0.0012.80 (1.63–4.81), P=0.001 Age (1 year)1.06 (1.04–1.08), P<0.0011.04 (1.02–1.06), P<0.001 Sex (0=female;1=male)2.11 (1.33–3.35), P=0.0011.45 (1.00–2.27), P=0.05ADMA, asymmetrical dimethylarginine.a Reference group. Open table in a new tab ADMA, asymmetrical dimethylarginine. After data adjustment for age and sex, the hazard ratios for death and CV events of plasma ADMA >75th percentile became higher (3.11 and 2.80, respectively) than the corresponding unadjusted (crude) estimates (2.43 and 2.41, respectively) and these results were little influenced by further adjustment for a large series of potential confounders (data not shown).6.Zoccali C. Bode-Böger S. Mallamaci F. et al.Plasma concentration of asymmetrical dimethylarginine and mortality in patients with end-stage renal disease: a prospective study.Lancet. 2001; 358: 2113-2117Abstract Full Text Full Text PDF PubMed Scopus (966) Google Scholar In this case, age and sex engendered ‘negative confounding’, that is the confounding effect of these variables resulted in an underestimation of the true hazard ratio of plasma ADMA. The problem of causality can rarely, if ever, be resolved only on the basis of well performed epidemiological studies or just on the basis of biological experiments. Indeed, establishing the nature of a given relationship (causal vs non causal) is a multidimensional process demanding not only accordance with a large a series of criteria (Table 5). Rothman and Greenland7.Rothman K.J. Greenland S. Causation and causal inference in epidemiology.Am J Public Health. 2005; 95: S144-S150Crossref PubMed Scopus (751) Google Scholar nicely illustrate the imperfection of these criteria considered in an isolated manner and carefully emphasize the importance of detailed knowledge on pathophysiological pathways in the assessment of causality.Table 5Hill's criteria for the inference of causalityStrength and the consistency of the observed relationshipSpecificity of the observed linkTemporality (the cause should always precede the effect)Biological gradient (a given relationship is more likely to underlie a causal effect in the presence of a dose-response relationship)CoherenceUnequivocal experimental evidence Open table in a new tab In summary, the problem of causality in biomedical research almost always demands application of various kind of studies, that is experimental and observational studies as well. Multivariate modeling of results of well-performed epidemiological studies provides valuable information for assessing the potential causal role of putative risk factors in human diseases. As previously discussed, in etiological research we aim to establish if a presumed causal relationship between a given risk factor and a given clinical outcome is unconfounded and therefore we pay careful attention to excluding the influence of confounding factors on the link being investigated. In contrast with etiological studies where we try to control for all possible confounding factors, in prognostic research we simply aim at establishing the probability of a given outcome and whether a given risk factor improves our prediction. As a consequence, the nature (causal or non causal) of risk factors being considered is not at issue. Left ventricular mass depends on a variety of risk factors like blood pressure, body mass, hemoglobin concentration, sympathetic activity, and other factors. Independently of the causes that may generate it, left ventricular hypertrophy is considered as one of the strongest predictors of death and CV complications and it is formally recommended by current guidelines for risk assessment refinement in hypertensive patients. To test whether left ventricular hypertrophy is useful for prognosis in the dialysis population, a prospective cohort study in 254 dialysis patients was performed.8.Zoccali C. Benedetto F.A. Mallamaci F. et al.Prognostic impact of the indexation of left ventricular mass in patients undergoing dialysis.J Am Soc Nephrol. 2001; 12: 2768-2774PubMed Google Scholar On crude (unadjusted) analysis a 10 g increase in left ventricular mass predicted a 45% increase in the risk of death (Table 6). To assess whether LV mass adds prognostic power for death to standard risk factors, investigators built up a multivariate Cox model including significant death predictors in the study population (that is age, gender, cholesterol, albumin, previous CV events, and C-reactive protein (CRP)) as well as left ventricular mass index (LVMI). In this model, LVMI maintained a significant prognostic value for all cause mortality and 10 g increase in LVMI entailed a 35% increase in the death risk (hazard ratio: 1.35, P<0.001; Table 6).Table 6Prognostic power of LVMI for all-cause mortality in dialysis patients8.Zoccali C. Benedetto F.A. Mallamaci F. et al.Prognostic impact of the indexation of left ventricular mass in patients undergoing dialysis.J Am Soc Nephrol. 2001; 12: 2768-2774PubMed Google ScholarUnits of increaseHazard ratio and 95% CIP-valueUnadjusted Cox regression analysis LVMI (g/height2.7)10 g/m2.71.45 (1.30–1.61)<0.001Multiple Cox regression analysis LVMI (g/height2.7)10 g/m2.71.35 (1.22–1.63)<0.001 Age1 year1.05 (1.02–1.07)<0.001 Gender0=female; 1=male2.46 (1.45–4.18)<0.001 Cholesterol1 mg/100 ml1.01 (1.00–1.01)0.002 Albumin1 g/100 ml0.41 (0.24–0.73)0.002 Previous CV events0=no; 1=yes1.83 (1.15–2.90)0.01 C-Reactive protein1 mg/l1.01 (1.00–1.01)0.04LVMI, left ventricular mass index; CV, cardiovascular. Open table in a new tab LVMI, left ventricular mass index; CV, cardiovascular. As LVMI added a significant predictive power above and beyond that provided by other risk factors (that is LVMI significantly improved the prediction of prognosis in ESRD patients), the conclusion of this paper was that measuring LVMI may be useful for risk stratification in the ESRD population. It is important noting that demonstrating that a given risk factor adds predictive power to established prognostic factors does not necessarily imply that the factor is useful in clinical practice.9.Manolio T. Novel risk markers and clinical practice.N Engl J Med. 2003; 349: 1587-1589Crossref PubMed Scopus (154) Google Scholar Whether a prognostic factor is useful depends on the magnitude of the additional prognostic power it conveys and, most importantly, on the demonstration that this information may improve the clinical decision process in every day clinical practice. Prognostic scores or risk calculators may be useful to refine prediction of clinical outcomes. For example, risk calculators based on Framingham risk factors are produced by the American and European society of Cardiology. The INdividual DAta aNAlysis of antihypertensive intervention trials (INDANA) risk calculator is a popular instrument for predicting the 5-years risk of CV death in adults with arterial hypertension (http://www.riskscore.org.uk).10.Pocock S.J. McCormack V. Gueyffier F. et al.A score for predicting risk of death from cardiovascular disease in adults with raised blood pressure, based on individual patient data from randomised controlled trials.BMJ. 2001; 323: 75-81Crossref PubMed Scopus (223) Google Scholar To build up this risk calculator, the INDANA investigators identified by Cox regression analysis a series of factors that were independently associated with the risk of CV death, that is age, sex, current cigarette smoking, systolic pressure, total cholesterol, creatinine, height, diabetes, left ventricular hypertrophy, previous myocardial infarction, and previous stroke. On the basis of Cox regression equation2.van Dijk P.C. Jager K.J. Zwinderman A.H. et al.The analysis of survival data in nephrology. Basic concepts and methods of Cox regression.Kidney Int. 2 July 2008Abstract Full Text Full Text PDF Scopus (59) Google Scholar: Ht=h0t*e(b*age+b1*sex+b2*smoking⋯etc) a risk calculator was built up. This is a typical prognostic model because it includes all factors that are predictive without bothering whether they satisfy the criteria for confounding or not. The application of this equation allows the calculation of the individual risk by the simple introduction into the same equation of the data defining the risk factors profile, (that is age, sex, blood pressure, and others) of the person wherein the risk estimate is being made. For example the 5-year probability of CV death in an individual with a risk profile detailed in Table 7 is 15.7%.Table 7INDANA risk calculatorAge56 yearsSexMaleCurrent cigarette smokerYesSystolic blood pressure170 mm HgTotal cholesterol250 mg/100 mlCreatinine, if known1.4 mg/100 mlHeight170 cmsDoes the patient have diabetes?NoDoes the patient have left ventricular hypertrophy?YesHas the patient already had a myocardial infarction?NoHas the patient already had a stroke?Yes Open table in a new tab This is truly a high risk because it is six times higher than that of a man of the same age without excess risk (2.6%). This calculator is useful not only to formulate a prognosis but also for patient education to reinforce treatment recommendations. If a man with similar characteristics lowers his blood pressure and stops smoking, his risk is expected to decrease substantially. The choice of the appropriate statistical modeling (etiological vs prognostic model) is strictly dependent on the research question being addressed and has important consequences on model building, data analysis and data interpretation. In etiological research control for confounding is fundamental. In prognostic research the only thing that matters is the accuracy of the prediction and therefore in this type of research the nature (causal or non causal) of risk factors being considered is irrelevant." @default.
• W1972774340 created "2016-06-24" @default.
• W1972774340 creator A5009797638 @default.
• W1972774340 creator A5046569845 @default.
• W1972774340 creator A5051761992 @default.
• W1972774340 creator A5086997559 @default.
• W1972774340 date "2008-12-01" @default.
• W1972774340 modified "2023-10-13" @default.
• W1972774340 title "Testing for causality and prognosis: etiological and prognostic models" @default.
• W1972774340 cites W1976541543 @default.
• W1972774340 cites W1992206451 @default.
• W1972774340 cites W2000480518 @default.
• W1972774340 cites W2011493627 @default.
• W1972774340 cites W2011804598 @default.
• W1972774340 cites W2058206695 @default.
• W1972774340 cites W2122019020 @default.
• W1972774340 cites W2148531805 @default.
• W1972774340 cites W3121647782 @default.
• W1972774340 doi "https://doi.org/10.1038/ki.2008.416" @default.
• W1972774340 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/18716602" @default.
• W1972774340 hasPublicationYear "2008" @default.
• W1972774340 type Work @default.
• W1972774340 sameAs 1972774340 @default.
• W1972774340 citedByCount "50" @default.
• W1972774340 countsByYear W19727743402012 @default.
• W1972774340 countsByYear W19727743402013 @default.
• W1972774340 countsByYear W19727743402014 @default.
• W1972774340 countsByYear W19727743402015 @default.
• W1972774340 countsByYear W19727743402016 @default.
• W1972774340 countsByYear W19727743402017 @default.
• W1972774340 countsByYear W19727743402018 @default.
• W1972774340 countsByYear W19727743402019 @default.
• W1972774340 countsByYear W19727743402020 @default.
• W1972774340 countsByYear W19727743402021 @default.
• W1972774340 countsByYear W19727743402022 @default.
• W1972774340 countsByYear W19727743402023 @default.
• W1972774340 crossrefType "journal-article" @default.
• W1972774340 hasAuthorship W1972774340A5009797638 @default.
• W1972774340 hasAuthorship W1972774340A5046569845 @default.
• W1972774340 hasAuthorship W1972774340A5051761992 @default.
• W1972774340 hasAuthorship W1972774340A5086997559 @default.
• W1972774340 hasBestOaLocation W19727743401 @default.
• W1972774340 hasConcept C121332964 @default.
• W1972774340 hasConcept C126322002 @default.
• W1972774340 hasConcept C137627325 @default.
• W1972774340 hasConcept C177713679 @default.
• W1972774340 hasConcept C62520636 @default.
• W1972774340 hasConcept C64357122 @default.
• W1972774340 hasConcept C71924100 @default.
• W1972774340 hasConceptScore W1972774340C121332964 @default.
• W1972774340 hasConceptScore W1972774340C126322002 @default.
• W1972774340 hasConceptScore W1972774340C137627325 @default.
• W1972774340 hasConceptScore W1972774340C177713679 @default.
• W1972774340 hasConceptScore W1972774340C62520636 @default.
• W1972774340 hasConceptScore W1972774340C64357122 @default.
• W1972774340 hasConceptScore W1972774340C71924100 @default.
• W1972774340 hasIssue "12" @default.
• W1972774340 hasLocation W19727743401 @default.
• W1972774340 hasLocation W19727743402 @default.
• W1972774340 hasOpenAccess W1972774340 @default.
• W1972774340 hasPrimaryLocation W19727743401 @default.
• W1972774340 hasRelatedWork W2008908422 @default.
• W1972774340 hasRelatedWork W208399605 @default.
• W1972774340 hasRelatedWork W2377142024 @default.
• W1972774340 hasRelatedWork W2412965123 @default.
• W1972774340 hasRelatedWork W2465854091 @default.
• W1972774340 hasRelatedWork W2902959513 @default.
• W1972774340 hasRelatedWork W3002986063 @default.
• W1972774340 hasRelatedWork W4237982933 @default.
• W1972774340 hasRelatedWork W4322705554 @default.
• W1972774340 hasRelatedWork W4322722809 @default.
• W1972774340 hasVolume "74" @default.
• W1972774340 isParatext "false" @default.
• W1972774340 isRetracted "false" @default.
• W1972774340 magId "1972774340" @default.
• W1972774340 workType "article" @default.