Marianthi Markatou

Active Computerized Pharmacovigilance Using Natural Language Processing, Statistics, and Electronic Health Records: A Feasibility Study

OBJECTIVE It is vital to detect the full safety profile of a drug throughout its market life. Current pharmacovigilance systems still have substantial limitations, however. The objective of our work is to demonstrate the feasibility of using natural language processing (NLP), the comprehensive Electronic Health Record (EHR), and association statistics for pharmacovigilance purposes. DESIGN Narrative discharge summaries were collected from the Clinical Information System at New York Presbyterian Hospital (NYPH). MedLEE, an NLP system, was applied to the collection to identify medication events and entities which could be potential adverse drug events (ADEs). Co-occurrence statistics with adjusted volume tests were used to detect associations between the two types of entities, to calculate the strengths of the associations, and to determine their cutoff thresholds. Seven drugs/drug classes (ibuprofen, morphine, warfarin, bupropion, paroxetine, rosiglitazone, ACE inhibitors) with known ADEs were selected to evaluate the system. RESULTS One hundred thirty-two potential ADEs were found to be associated with the 7 drugs. Overall recall and precision were 0.75 and 0.31 for known ADEs respectively. Importantly, qualitative evaluation using historic roll back design suggested that novel ADEs could be detected using our system. CONCLUSIONS This study provides a framework for the development of active, high-throughput and prospective systems which could potentially unveil drug safety profiles throughout their entire market life. Our results demonstrate that the framework is feasible although there are some challenging issues. To the best of our knowledge, this is the first study using comprehensive unstructured data from the EHR for pharmacovigilance.

Intrahepatic levels of CXCR3‐associated chemokines correlate with liver inflammation and fibrosis in chronic hepatitis C

Chemokines, chemotactic cytokines, may promote hepatic inflammation in chronic hepatitis C virus (HCV) infection through the recruitment of lymphocytes to the liver parenchyma. We evaluated the association between inflammation and fibrosis and CXCR3‐associated chemokines, interferon‐γ (IFN‐γ)–inducible protein 10 (IP‐10/CXCL10), monokine induced by IFN‐γ (Mig/CXCL9), and interferon‐inducible T cell α chemoattractant (I‐TAC/CXCL11), in HCV infection. Intrahepatic mRNA expression of these chemokines was analyzed in 106 chronic HCV‐infected patients by real‐time PCR. The intrahepatic localization of chemokine producer cells and CXCR3+ lymphocytes was determined in selected patients by immunohistochemistry. We found elevated intrahepatic mRNA expression of all three chemokines, most markedly CXCL10, in chronic HCV‐infected patients with higher necroinflammation and fibrosis. By multivariable multivariate analysis, intrahepatic CXCL10 mRNA expression levels were significantly associated with lobular necroinflammatory grade and HCV genotype 1. In the lobular region, CXCL10‐expressing and CXCL9‐expressing hepatocytes predominated in areas with necroinflammation. Strong CXCL11 expression was observed in almost all portal tracts, whereas CXCL9 expression varied considerably among portal tracts in the same individual. Most intrahepatic lymphocytes express the CXCR3 receptor, and the number of CXCR3+ lymphocytes was increased in patients with advanced necroinflammation. Conclusion: These findings suggest that the CXCR3‐associated chemokines, particularly CXCL10, may play an important role in the development of necroinflammation and fibrosis in the liver parenchyma in chronic HCV infection. (Hepatology 2008.)

Automated Acquisition of Disease–Drug Knowledge from Biomedical and Clinical Documents: An Initial Study

OBJECTIVE Explore the automated acquisition of knowledge in biomedical and clinical documents using text mining and statistical techniques to identify disease-drug associations. DESIGN Biomedical literature and clinical narratives from the patient record were mined to gather knowledge about disease-drug associations. Two NLP systems, BioMedLEE and MedLEE, were applied to Medline articles and discharge summaries, respectively. Disease and drug entities were identified using the NLP systems in addition to MeSH annotations for the Medline articles. Focusing on eight diseases, co-occurrence statistics were applied to compute and evaluate the strength of association between each disease and relevant drugs. RESULTS Ranked lists of disease-drug pairs were generated and cutoffs calculated for identifying stronger associations among these pairs for further analysis. Differences and similarities between the text sources (i.e., biomedical literature and patient record) and annotations (i.e., MeSH and NLP-extracted UMLS concepts) with regards to disease-drug knowledge were observed. CONCLUSION This paper presents a method for acquiring disease-specific knowledge and a feasibility study of the method. The method is based on applying a combination of NLP and statistical techniques to both biomedical and clinical documents. The approach enabled extraction of knowledge about the drugs clinicians are using for patients with specific diseases based on the patient record, while it is also acquired knowledge of drugs frequently involved in controlled trials for those same diseases. In comparing the disease-drug associations, we found the results to be appropriate: the two text sources contained consistent as well as complementary knowledge, and manual review of the top five disease-drug associations by a medical expert supported their correctness across the diseases.

Semiparametric Estimation of Regression Models for Panel Data

Linear models with error components are widely used to analyze panel data. Some applications of these models require knowledge of the probability densities of the error components. Existing methods handle this requirement by assuming that the densities belong to known parametric families of distributions (typically the normal distribution). This paper shows how to carry out nonparametric estimation of the densities of the error components, thereby avoiding the assumption that the densities belong to known parametric families. The nonparametric estimators are applied to an earnings model using data from the Current Population Survey. The models transitory error component is not normally distributed. Use of the nonparametric density estimators yields estimates of the probability that individuals with low earnings will become high earners in the future that are much lower than the estimates obtained under the assumption of normally distributed error components. JEL Classification: C13, C14, C23

Peripheral CXCR3-Associated Chemokines as Biomarkers of Fibrosis in Chronic Hepatitis C Virus Infection

BACKGROUND CXCR3-associated chemokines CXCL9-CXCL11 promote histologic progression in chronic hepatitis C virus (HCV) infection, as indicated by elevated intrahepatic levels of messenger RNA in patients with advanced inflammation and fibrosis. We evaluated the potential of peripheral chemokine levels to discriminate among patients with chronic HCV infection who had different stages of fibrosis. METHODS Peripheral levels of CXCR3-associated chemokines were measured by enzyme-linked immunosorbent assay of plasma samples obtained from 93 patients with chronic HCV infection. Of the subjects, 79 (85%) were white, and 68 (73%) were infected with HCV genotype 1. RESULTS Expression of all 3 chemokines, when analyzed as a group, was significantly associated with intrahepatic inflammation and fibrosis. Plasma levels of CXCL10 were significantly elevated in patients with advanced fibrosis, whereas CXCL9 levels were significantly elevated in patients with advanced inflammation. By proportional odds multivariate modeling, we observed an association between fibrosis and CXCL10 (P< .002) as well as between fibrosis and inflammation (P<.001). Of the individual parameters, the CXCL10 level was most useful in identifying patients with more-severe (stage 3-4) fibrosis. Discriminatory ability was improved by the combination of CXCL10 and CXCL9. CONCLUSIONS The strong association between CXCR3-associated chemokines and fibrosis suggests that they may have promise as noninvasive markers of hepatic fibrosis in a predominantly white HCV genotype 1-infected population.

Weighted likelihood equations with bootstrap root search

Abstract We discuss a method of weighting likelihood equations with the aim of obtaining fully efficient and robust estimators. We discuss the case of continuous probability models using unimodal weighting functions. These weighting functions downweight observations that are inconsistent with the assumed model. At the true model, therefore, the proposed estimating equations behave like the ordinary likelihood equations. We investigate the number of solutions of the estimating equations via a bootstrap root search; the estimators obtained are consistent and asymptotically normal and have desirable robustness properties. An extensive simulation study and real data examples illustrate the operating characteristics of the proposed methodology.

Pharmacodynamics of PEG‐IFN α differentiate HIV/HCV coinfected sustained virological responders from nonresponders

Pegylated interferon (PEG‐IFN) has become standard therapy for hepatitis C virus (HCV) infection. We evaluated whether PEG‐IFN pharmacodynamics and pharmacokinetics account for differences in treatment outcome and whether these parameters might be predictors of therapeutic outcome. Twenty‐four IFN‐naïve, HCV/human immunodeficiency virus–coinfected patients received PEG‐IFN α‐2b (1.5 μg/kg) once weekly plus daily ribavirin (1,000 or 1,200 mg) for up to 48 weeks. HCV RNA and PEG‐IFN α concentrations were obtained from samples collected frequently after the first 3 PEG‐IFN doses. We modeled HCV kinetics incorporating pharmacokinetic and pharmacodynamic parameters. Although PEG‐IFN concentrations and pharmacokinetic parameters were similar in sustained virological responders (SVRs) and nonresponders (NRs), the PEG‐IFN α‐2b concentration that decreases HCV production by 50% (EC50) was lower in SVRs compared with NRs (0.04 vs. 0.45 μg/L [P = .014]). Additionally, the median therapeutic quotient (i.e., the ratio between average PEG‐IFN concentration and EC50[C̄/EC50]), and the PEG‐IFN concentration at day 7 divided by EC50 (C(7)/EC50) were significantly increased in SVRs compared with NRs after the first (10.1 vs. 1.0 [P = .012], 2.8 vs. 0.3 [P = .007], respectively) and second (14.0 vs. 1.1 [P = .016], 5.4 vs. 0.4 [P = .02], respectively) PEG‐IFN doses. All 3 parameters may be used to identify NRs. In conclusion, PEG‐IFN concentrations and pharmacokinetic parameters do not differ between SVRs and NRs. In contrast, pharmacodynamic measurements—namely EC50, the therapeutic quotient, and C(7)/EC50—are different in coinfected SVRs and NRs. These parameters might be useful predictors of treatment outcome during the first month of therapy. (HEPATOLOGY 2006;43:943–953.)

Bounded Influence and High Breakdown Point Testing Procedures in Linear Models

Abstract Three classes of testing procedures based on one-step high breakdown point bounded influence estimators, for testing subhypotheses in linear models are developed. These are drop-in-dispersion, Wald-type, and score-type tests. The asymptotic distributions of these testing procedures are obtained under the null hypothesis and under contiguous alternatives. Their stability properties are studied in terms of their influence functions and breakdown points. It is shown that the tests have bounded influence functions. For the Wald-type tests, the level and power breakdowns are determined by the breakdown point of the parameter estimate and the associated variance-covariance matrix. The drop-in-dispersion test exhibits high-level breakdown but not high power breakdown point. Similar behavior is exhibited by the score-type tests. But slight modifications can be made in the construction of the test statistics to ensure high breakdown points in terms of both level and power. An example is given to illustrat...

Gene symbol disambiguation using knowledge-based profiles

MOTIVATION The ambiguity of biomedical entities, particularly of gene symbols, is a big challenge for text-mining systems in the biomedical domain. Existing knowledge sources, such as Entrez Gene and the MEDLINE database, contain information concerning the characteristics of a particular gene that could be used to disambiguate gene symbols. RESULTS For each gene, we create a profile with different types of information automatically extracted from related MEDLINE abstracts and readily available annotated knowledge sources. We apply the gene profiles to the disambiguation task via an information retrieval method, which ranks the similarity scores between the context where the ambiguous gene is mentioned, and candidate gene profiles. The gene profile with the highest similarity score is then chosen as the correct sense. We evaluated the method on three automatically generated testing sets of mouse, fly and yeast organisms, respectively. The method achieved the highest precision of 93.9% for the mouse, 77.8% for the fly and 89.5% for the yeast. AVAILABILITY The testing data sets and disambiguation programs are available at http://www.dbmi.columbia.edu/~hux7002/gsd2006

Weighted likelihood estimating equations: The discrete case with applications to logistic regression

Abstract We discuss a method of weighting the likelihood equations with the aim of obtaining fully efficient and robust estimators. We discuss the case of discrete probability models using several weighting functions. If the weight functions generate increasing residual adjustment functions then the method provides a link between the maximum likelihood score equations and minimum disparity estimation, as well as a set of diagnostic weights and a goodness of fit criterion. However, when the weights do not generate increasing residual adjustment functions a selection criterion is needed to obtain the robust root. The weight functions discussed in this paper do not automatically downweight a proportion of the data; an observation is significantly downweighted only if it is inconsistent with the assumed model. At the true model, therefore, the proposed estimating equations behave like the ordinary likelihood equations. We apply our results to several discrete models; in addition, a toxicology experiment illustrates the method in the context of logistic regression.