Treat or test first? Decision analysis of empirical antiviral treatment of influenza virus infection versus treatment based on rapid test results

Abstract

Background: neuraminidase (NA) inhibitors have recently become available for treatment of influenza. Rapid antigen detection assays at ‘point-of-care’ may improve the accuracy of clinical diagnosis, but the value of these techniques in assisting with the appropriate use of antivirals remains controversial. Objective: to compare the diagnostic utilities of two management strategies for influenza, empirical antiviral therapy versus therapy based on a positive rapid test result in pre-epidemic and epidemic periods. Study design: a threshold decision analytic model was designed to compare these competing strategies and sensitivity analysis performed to examine the impact of diagnostic variables on the expected utility of the decision with a range of prior probabilities of infection between 1 and 50%. Results: on the basis of the calculated sensitivity (77%) and specificity (95%) of a point-of-care test for influenza, pre-treatment testing was preferred and cost-effective in non-epidemic stage of the influenza cycle. The alternative strategy of empirical treatment produces a higher utility value during epidemics, but may result in overuse of antivirals for low-risk populations. The two strategies had equivalent efficacy when the probability of influenza was 42%. Conclusions: Patients with flu-like illness, who present outside the influenza outbreak and are considered to be at low risk for influenza-related complications, should be tested to confirm the diagnosis before starting antiviral treatment with a NA inhibitor. The most important variables in the model were the accuracy of the clinical diagnosis and the pre-test probability of influenza. A threshold probability of influenza of 42% would dictate changing from the rapid testing strategy to a ‘treat regardless’ strategy.

Introduction

New neuraminidase (NA) inhibitors with broad-spectrum activity against all nine influenza A subtypes and influenza B have recently become available for treatment and prophylaxis of influenza. Inhaled zanamivir (Relenza®) is now licensed in Northern America, Europe and Australia and oral oseltamivir (Tamiflu®) has been approved in Canada, Switzerland, and USA. They can significantly reduce the severity and duration of symptoms, time taken to return to normal activities and use of antibiotics and relief medications in both the general population and in patients at high risk for complications (Hayden et al., 1997, Hayden et al., 1999, Monto et al., 1999a, Gubareva et al., 2000). However, for therapy to be effective it needs to be commenced within 36–48 h of onset of illness.

Influenza virus infections typically occur in outbreaks lasting 8–12 weeks during winter, and are uncommon outside these periods. Diagnosis of influenza on clinical grounds alone is both insensitive and non-specific in non-epidemic situations (Carrat et al., 1997, Nicholson et al., 1998). Although cell culture remains the gold standard for the laboratory diagnosis and surveillance of influenza, it has limited clinical utility because results are obtained too late for effective intervention. Rapid diagnostic assays for influenza have demonstrated high specificity, but poor sensitivity compared with cell culture (Johnston and Seigel, 1991, Doing et al., 1998) and their role in the diagnosis of influenza and initiation of treatment remains unclear. It is expected that by combining the clinical diagnosis of influenza with office-based rapid testing, candidates for antiviral treatment will be better selected.

The purpose of this study was to compare two strategies for management of clinical influenza using new data on the performance of commercial rapid tests and efficacy of influenza treatment. Decision analysis is a popular tool to structure a clinical problem and to identify the main determinants of diagnostic and therapeutic choice (Lilford et al., 1998). It uses Bayesian probabilities together with values assigned to different outcomes to determine the best course of action. We chose a threshold model of decision analysis to compare the value of empirical treatment with the treatment of clinical influenza based on rapid test results.