Signal Detection Methodology @ Lambda
02 Feb 2018
Introduction:
Following its inception in 2009, Pharmacovigilance at Lambda has grown organically and signal management activities are conducted by Lambda PV since 2010. Initially, signal detection was performed based on vol. 9. However, current processes (from 2012) are based on new signal detection module IX under GVP guideline, along with an additional guidance taken from CIOMS VIII. Both the process as well as the expertise on signal detection have evolved over time. Lambda is capable of carrying out not only the traditional (qualitative) methods for detecting signals but also have equal calibre on statistical (quantitative) methods. Varied procedures that are prevalent within Lambda to carry out qualitative analysis have been well accepted and acknowledged by regulatory authorities.
Experienced PV Medical Reviewers are the experts who carry out the signal detection and monitoring activity at Lambda, whereby the products for signal management are allocated as per therapeutic areas. The periodicity of a product is assessed based on the safety profile so far and defined within the signal management schedule. Determination of the periodicity is based on the following key safety profiling factors of the product like: pattern, seriousness & severity, frequency of events that are under monitoring, PSUR cycle or any special recommendation from the regulatory authority.
Signal Detection Methodology:
By and large, Lambda relies on an integrated approach using either qualitative or quantitative data mining algorithms for identification of potential signals. However identified potential signals through quantitative methods are reinforced by qualitative assessment of each of the statistically identified signals to validate a true signal.
Traditional (Qualitative) Methods:
Most of the organizations weigh their choices on signal detection methodology primarily based on the amount of safety information that has been gathered for the product under evaluation. Hence, most of our Marketing Authorization Holders (MAHs) having low volume of collected safety reports, but with well documented cases presenting DMEs (designated medical events), TMEs (targeted Medical events) for assessment, often have opted for traditional signal detection methodology that uses ‘Index case’/’Striking case’ system.
Qualitative signal detection method will be performed based on an overall evaluation of the product’s profile keeping in lieu the key favorable factors like completeness of the available data, event-drug relatedness, strength of the causal relationship of the product with the adverse reactions, specificity of the causal association, and objective data based evidence for identified event (event objectivity) and it’s frequency. The focal point and a key factor that determines effective identification of potential signals is the clinical acumen of the Medical reviewer and his familiarity with the product’s behavior so far, that would enable identification of potential signals effectively from either single cases or from available case series.
Quantitative Data Mining Algorithms (DMA):
For products with large safety datasets, besides traditional methodology, we also perform Quantitative Signal Detection, using an inbuilt data mining method within our safety database.
For an MAH’s product, the safety database (PvNET) condenses large and complex dataset into 2×2 contingency tables for analysis of signals of disproportionate reporting that provides a platform for identification of Drug Event combinations (DEC) in terms of PRR (Proportionate Reporting Rate). PRR (+) & PRR (-) provides Signals of disproportionate reporting (SDR) respective upper and lower limits on CIs (Confidence Intervals); generally 95% CI is considered. Other methods while evaluating the PRR in the 2×2 contingency table can be compounded with Chi square values. Based on the above outputs, SDR are identified depending on the threshold values entered at the start of quantitative SD activity. In addition to identifying DEC at LLT or PT level, our database (PvNET) also allows data mining at Standardized MedDRA Queries (SMQ) level.
The SDRs thus identified, are further evaluated clinically, using a functionality that is inbuilt within PvNET for Qualitative signal detection that allows appropriate segregation of true clinical signals from those of the “statistical noise”.
Further Signal Management:
Validation of signal:
Following the identification of a potential signal, validation of the signal is initiated. Validation of a signal is done through varied mechanisms that includes comprehensive analysis of the case, referring to the Innovator’s label or standard textbooks to see whether the event is already a known fact, conducting a standalone literature search for the drug-event combination, or if required, screening through larger database/s like those of Eudravigilance. Based on the finality from the validation exercise of the signal, further actions are planned for further signal management.
In case, the potential signal fails in the validation process, the signal is tagged as a ‘refuted signal’, whereby the events are logged in corresponding identified signal spreadsheet which would be tracked and monitored in future to assess similar or newer safety issues highlighted during that monitoring phase. Based on conclusive assessed evidence on validated signals, such validated signals are discussed at the Safety Review Meetings that define and design further action plan for drafting and implementation.
Prioritization of Signals:
Signal prioritization depends on the probability of the exposure and the impact the event might have on the patients that are being treated with the medicinal product; and points like severity of the event, its reversibility, potential scope for prevention, along with its possible clinical outcome are taken into consideration. During evaluation and review of the validated signals for prioritization, supreme priority is given towards assessing and communicating the health impacts on the special populations such as pregnant females, paediatrics and the elderly, however not limiting to the overall size of the general population at risk.
The validated signal is further assessed to identify on the need for any additional data collection and/or regulatory action. Additional data are gathered with the use of SMQs, whereby the product’s class effects or its potential complications are reviewed. In case of any recommendations following the signal assessment, these are discussed with the MAHs, who in turn, upon their final judgement or conclusive evidence, share the information with the regulatory authorities.
Conclusion:
Identification, tracking and monitoring of signals can be a herculean task, however we at Lambda have initiated signal management activities and are continuously evolving on this process. The guiding bodies with their guidelines and a robust data mining tool is what we have; we definitely are looking forward for further improvisations in this field to create an edge in the competitive landscape.
References: GVP Module IX and CIOMS VIII.