Exploring New Zealand prescription data using sequence symmetry analyses for predicting adverse drug reactions. (13th December 2016)

• Record Type: Journal Article
• Authors: Nishtala, P. S.; Chyou, T.‐y.
• Is Part Of: Journal of clinical pharmacy and therapeutics. Volume 42:Number 2(2017)
• Page Start: 189
• Page End: 194
• Publication Date: 2016-12-13
• Subjects: adverse drug reaction -- case crossover -- pharmacoepidemiology -- prescriptions -- sequence symmetry analyses; Clinical pharmacology -- Periodicals; Chemotherapy -- Periodicals; 615
• DOI: 10.1111/jcpt.12491
• Languages: English
• ISSNs: 0269-4727

Abstract

Summary: What is known and objective: Prescription sequence symmetry analyses (PSSA) is a ubiquitous tool employed in pharmacoepidemiological research to predict adverse drug reactions (ADRs). Several studies have reported the advantage of PSSA as a method that can be applied to a large prescription database with computational ease. The objective of this study was to validate New Zealand (NZ) prescription database as a potential source for identifying ADRs using the PSSA method. Methods: We analysed de‐identified individual‐level prescription data for people aged 65 years and above for the period 2005 to 2014 from the pharmaceutical collections supplied by the NZ Ministry of Health. We selected six positive controls that have been previously investigated and reported for causing ADRs. The six positive controls identified were amiodarone (repeated twice), frusemide, simvastatin, lithium and fluticasone. Amiodarone and lithium have been reported to induce thyroid dysfunction. Simvastatin reported to cause muscle cramps while fluticasone is well documented to cause oral candidiasis. Thyroxine was identified as a marker drug to treat hypothyroidism associated with amiodarone and lithium. Carbimazole was identified as a marker drug to treat hyperthyroidism associated with amiodarone use. Quinine sulphate was identified as a marker drug to treat muscle cramps associated with statins. In addition, we also analysed six negative controls that are unlikely to be associated with ADRs. The main outcome measure is to determine associations with ADRs using adjusted sequence ratios (ASR), and 95% confidence intervals Results and discussion: Our analyses confirmed a significant signal for all six positive controls. Significant positive associations were noted for amiodarone [ASR = 3·57, 95% CI (3·17–4·02)], and lithium chloride induced hypothyroidism [ASR = 3·43, 95% CI (2·55–4·70)]. Amiodarone was also strongly associated with hyperthyroidism [ASR = 8·81 95% CI (5·86–13·77)]. Simvastatin was associated with muscle cramps [ASR = 1·69, 95% CI (1·61–1·77)]. Fluticasone was positively associated with oral candidiasis [ASR = 2·34, 95% CI (2·19–2·50)]. Frusemide was associated with hypokalaemia [ASR = 2·94, 95% CI (2·83–3·05]). No strong associations were noted for the negative pairs. It is important to highlight that PSSA automatically controls for all confounding factors including unknown and unmeasured confounding variables, plus the effect of temporal trend in prescriptions, and hence allows a more robust ADR detection especially when confounding factors are difficult to determine or measure. What is new and conclusion: New Zealand prescription database can be a potential source to identify ADRs engaging the PSSA method, and this could complement pharmacovigilance surveillance in NZ. The PSSA can be an important method for post‐marketing surveillance and monitoring of ADRs which have relatively short latency. However, the predictive validity of PSSA will be compromised in certain scenarios, particularly when sample size is small, when new drugs are in the market and data are sparse. Abstract : Histograms demonstrate the asymmetries in marker‐drug initiation times, as a result of the ADRs caused by the index‐drug. In each histogram, each bin represents a 30‐day time interval, an din addition, individuals using the index‐drug and the corresponding marker‐drug on the same day are excluded.