DETECTION OF DE-CHALLENGE IN SPONTANEOUS REPORTING SYSTEMS

INTRODUCTION

Causality assessment (CA) is a method used in pharmacovigilance to evaluate the relationship between drugs and reported adverse drug reactions (ADR). It involves identifying the drugs causing ADR and evaluating the reaction of dechallenge. Dechallenge is a response observed in a patient, such as a reduction or disappearance of ADR after drug withdrawal. There are two types: positive dechallenge, which resolves with drug withdrawal, and negative dechallenge, which follows its own course. Rechallenge is essential to confirm the cause and relationship of ADR. Prescribers and patients may not follow typical procedures for rechallenge, and prescribers prefer dechallenge over rechallenge.

Data mining is a statistical approach used to discover useful patterns from large amounts of data, including algorithms for classification, prediction, clustering, and association. The Multi-item Gamma Position Shrinker is used by the Food and Drug Administration (FDA) in the US for signalling potential ADRs. Naive Bayes (NB) is a statistical classifier used for classification, but it assumes mutually independent variables. This method can lead to a simple prediction framework, but there is a possibility of zero probability. To overcome this issue, a standard technique like Laplacian correction was used.

 

MATERIALS AND METHODS:

The Medical Dictionary for Regulatory Activities (MedDRA) is an international medical terminology used to describe adverse events at five levels: System Organ Class (SOC), High Level Group Term (HLGT), Higher Level Term (HLT), Preferred Term (PT), and Lowest Level Term (LLT). In a study involving FDA ADRs from 2011 and 2012, data was extracted from the FDA's SRS, and duplicate reports were deleted. The FDA stored diseases category at the Preferred Terms (PT) level. Researchers suggested using a coarser-grained adverse event representation, SOC, than the PT level for data mining.

Data were loaded from the FDA's text file into an Oracle database using Extract, Transform, and Load (ETL) tools, and indications were constructed using patient identifiers. Records with SOC were considered for dechallenge classification. The attributes considered for dechallenge included disease categories, valid trade and verbatim names, outcomes like life-threatening (LT), death (DE), congenital anomaly (CA), hospitalization-initial or prolonged (HO), disability (DS), and required intervention (RI) to prevent permanent impairment.

 

The performance of the data mining algorithm was estimated using parameters like accuracy, error, precision, and Receiver Operating Characteristic (ROC) curve. Accuracy, precision, and error were calculated using Formulas 1, 2, and 3. The ROC curve was used to display the graphical representation of the perfect, liberal, random, and conservative performance of an algorithm.

                                     Formula 1:

                                    

                                     Formula 2:

 

                                     Formula 3:

 

RELATED WORKS:

Statistical data mining techniques have been applied in post-marketing surveillance, with various studies examining safety signal detection problems, data sources, and methods used in safety signal classification. Decision tree algorithms have been applied for drug safety classification. The Naive Credal Classifier, an extension of NB, is used for high uncertain information domains. Denis et al. propose a classifier for eliminating noise in datasets. Chen and Shengrui propose classification methods for high-dimensional data. Kotsiantis and Pintelas modified NB classifiers using bagging and boosting procedures. Zhang et al. proposed a novel model called Hidden Naive Bayes to avoid computational complexity.

 

DATA MINING MODEL:

The data mining model identified patients by unique numbers and analyzed drug and disease category outcomes. FDA classified drugs as 1 for valid trade and 2 for verbatim name. Disease categories were denoted at the SOC level using FDA's PT and MedDRA's PT using xml mapping.

 

THE ALGORITHMS USED:

Naive bayas

The fundamental assumption to attribute independence was considered in this study. Dechallenge attribute presented in FDA records were taken as class label for detection. NB theorem given in Formula 4 had been used to calculate the probability of an outcome. The class label attribute dechallenge had two distinct values (Yes, No) represented by hypothesis (H).

                                   

 

P(H/X) is the posterior probability where hypothesis (H) represents the presence of dechallenge with X as known disease category, drug code and outcome. P (X/H) is the posterior probability of X about H. P (H) is the prior probability of H regardless of disease category, drug code and outcome. P(X) is the prior probability of X.

For calculating the prior probability P (X), dechallenge record sets with 'unknown' category were filtered. Then the posterior probability was calculated based on outcome, disease category and drug code. The data set of 2011 and 2012 records contained the constraints mentioned for the failure of Naive Bayes classifier.

 

IMPROVED NAIVE BAYES:

The algorithm calculates the Influence Factor for attributes like drug code, disease category, and outcome on the label dechallenge, using Formula 5 to determine the dependability of an attribute value on the class attribute.

 

 

Where I(X/Ci) =Influence Factor

N(X\C _i ) =Number of records in which attribute value X had the class label C _i and N(Ci)=Total Number of records in which the class label were C _i .The dataset was divided based on the class label dechallenge 'Yes' and 'No'. Influence factor for attributes with high values were taken and others ignored.

 

CONCLUSION:

The FDA and WHO's data lacks health science education and training for patient care. Post-marketing surveillance techniques like detecting dechallenge can help practitioners and prescribers understand drugs with reactions. Unknown samples should be properly classified for data analysis. FDA suggests dechallenge evidence as the most important criteria for causality reviews. Classification algorithms show that NB+ outperforms NB in accuracy and minimal error.

 

Student Name: Mohamed Salih Aslam

Student ID: 188/092023

Qualification: B.Pharmacy

e-Mail ID: msalihaslam@gmail.com