- Research article
- Open access
- Published:
Antiviral prophylaxis during pandemic influenza may increase drug resistance
BMC Infectious Diseases volume 9, Article number: 4 (2009)
Abstract
Background
Neuraminidase inhibitors (NI) and social distancing play a major role in plans to mitigate future influenza pandemics.
Methods
Using the freely available program InfluSim, the authors examine to what extent NI-treatment and prophylaxis promote the occurrence and transmission of a NI resistant strain.
Results
Under a basic reproduction number of R0 = 2.5, a NI resistant strain can only spread if its transmissibility (fitness) is at least 40% of the fitness of the drug-sensitive strain. Although NI drug resistance may emerge in treated patients in such a late state of their disease that passing on the newly developed resistant viruses is unlikely, resistant strains quickly become highly prevalent in the population if their fitness is high. Antiviral prophylaxis further increases the pressure on the drug-sensitive strain and favors the spread of resistant infections. The authors show scenarios where pre-exposure antiviral prophylaxis even increases the number of influenza cases and deaths.
Conclusion
If the fitness of a NI resistant pandemic strain is high, any use of prophylaxis may increase the number of hospitalizations and deaths in the population. The use of neuraminidase inhibitors should be restricted to the treatment of cases whereas prophylaxis should be reduced to an absolute minimum in that case.
Background
Neglecting the possible emergence of NI resistance, modeling studies have suggested that an influenza pandemic may be contained if treatment and prophylaxis are introduced immediately [1–3]. Switzerland has stockpiled sufficient NI to treat 25% of the population and considers using some of the stockpile for prophylaxis in health care workers and essential services (fire brigade, police, etc). As influenza viruses mutate constantly, widespread use of antivirals in the case of a pandemic could cause selection pressure which could lead to the emergence and spread of NI resistant strains. De novo emergence of a resistant strain does not necessarily cause negative outcomes; it is the transmission fitness of the resistant strain which plays a major role [4–7]. Simulation studies on HIV [8] and HSV-2 [9] have also shown the important role of transmissibility in driving the dynamics of drug resistance. Influenza resistant strains that show nearly the same transmission fitness as wild type strains have been detected [10]. Also, data from the European system VIRGIL show, that a NI resistant seasonal influenza A (H1N1) strain circulates in several countries [11]. NI resistance leads to treatment failures and thus, to an increased number of hospitalizations and deaths attributable to influenza. As antiviral prophylaxis and treatment are relevant for future interventions against influenza, understanding the implications of NI resistance and the dynamics of its spread is essential. We examine how NI treatment and prophylaxis contribute to the emergence and circulation of NI resistant strains and to changes in the epidemiology of infection.
Methods
We extend the freely available deterministic simulation program InfluSim by including the emergence and the spread of NI resistance [12]. InfluSim version 2.2 distinguishes drug sensitive and resistant infections, and allows for prophylaxis [13].
We employ a basic reproduction number of R 0 = 2.5 [14], and assume that one third of all infected individuals remain asymptomatic, that another third becomes moderately sick and that the remaining third becomes severely sick and seeks medical help (Fig. 1). A proportion of severely sick cases needs hospitalization and may die from the disease, depending on the age and on the risk group of the patient (see Additional file 1, Table A1). Adults who develop severe disease are sick and contagious for 7 days on average and need 5 more days to recover before they can resume work. Treatment reduces their remaining duration of the disease by 25%, their contagiousness by 80% [2] and their need of hospitalization (and concurrently their risk of death) by 50% [15]. A fraction of the population between 20 and 60 years of age is offered antiviral prophylaxis. We assume that people who take prophylaxis are less susceptible to the infection with the drug sensitive strain (the standard value of 50% is varied in an uncertainty analysis). If they are infected, twice as many of them remain asymptomatic than without prophylaxis; their contagiousness, their duration of being sick and their need for hospitalization are reduced in the same way as described above for therapeutic treatment. We furthermore assume that only half of the individuals who take prophylaxis become immune after asymptomatic infection whereas the other half is left susceptible (Fig. 1).
Simulations start with a single drug sensitive infection in a Swiss population of 100,000 inhabitants. We assume that 4.1% of children up to 12 years of age and 0.32% of teenagers and adults infected with the drug sensitive virus develop a NI resistant infection if they take antiviral drugs [16, 17]. The probability that they pass on the resistant virus to others is small, as on average their contagiousness has already decreased considerably before the resistant virus appears. This is because (i) patients seek medical treatment on average 24 hours after onset of symptoms, (ii) their contagiousness is assumed to decline exponentially over time with 90% of the total infectiousness being spent within the first half of the infectious period, and (iii) if de novo NI resistance occurs in a patient, it occurs at a random time point between treatment and the end of contagiousness. The resistant virus can have a reduced transmissibility (fitness) compared to the drug-sensitive strain, while having the same pathogenicity, causing the same natural history of disease and inducing the same degree of immunity, but cases infected with the NI resistant virus no longer respond to antiviral drugs. Simulations are performed for a variety of parameter values: The fitness of the NI resistant strain is varied from 80 to 100%. The percentage of adults between 20 and 60 years of age who receive prophylaxis is varied from 0 to 20%. Prophylaxis is assumed to reduce the susceptibility of the recipient by 50% (varied in an uncertainty analysis from 0 to 100%). Fifty percent of prophylactically treated individuals who experience asymptomatic infection become immune after clearing the infection (varied in an uncertainty analysis from 0 to 100%). We run simulations with and without prophylaxis, with and without resistance, and observe the total number of hospitalizations due to severe influenza and the total work loss for people who receive prophylaxis. No ethics committee was required to grant permission for this investigation.
Results
Simulation results
Figures 2a–c show how the prevalence of infection with the drug sensitive and the resistant virus change during the pandemic wave if all severe cases are treated with NI and if additionally 0%, 10% or 20% of the people between 20 and 60 years of age receive prophylaxis (assuming that the fitness of the NI resistant strain is as high as that of the drug-sensitive one, i.e. 100%). The simulations start with one drug-sensitive infection in a susceptible population of 100,000 individuals. NI resistance develops de novo and gradually builds up during the epidemic wave. If no prophylaxis is given, the drug-sensitive strain dominates most of the epidemic wave (full curve) and the resistant strain (dashed curve) only becomes prevalent in the end (Fig. 2a).
Prophylaxis increases the pressure on the drug sensitive strain and favors the transmission of the resistant strain. The prevalence of the NI resistant strain increases considerably if 10% of all people between 20 and 60 years of age are given prophylaxis (Fig. 2b). If the prophylaxis coverage is increased to 20%, the resistant strain very quickly dominates the epidemic wave and the majority of cases are infected with the resistant strain (Fig. 2c). With increasing prophylaxis coverage, the peak number of infected individuals increases from 17,000 to 25,000 and the total number of infected individuals increases from 62,000 to 72,000 (Figs. 2a–c).
Individuals who are infected with the NI resistant strain do not respond to antiviral treatment. Even without any prophylaxis in the population, 13.7% of all treated patients are treated in vain due to resistant infection. This fraction increases to 43.4% and 74.5%, respectively, if 10% or 20% receive prophylaxis.
The joint influence of antiviral prophylaxis and fitness of the resistant strain is shown in Figs. 3a–c. We assume that during the whole course of the epidemic 0 to 20% of people between 20 to 60 years of age receive prophylaxis. Fitness of the resistant virus is varied from 80 to 100% (simulations with lower fitness values yield results which are very similar to the 80% scenario). Antiviral prophylaxis increases the number of treatment failures in the population (Fig. 3a). If the fitness of the resistant strain is less than 80%, this effect may be regarded as negligible, but for higher fitness values and notably for a combination of high fitness and large prophylaxis coverage, the results become increasingly pessimistic. If the fitness of the resistant strain is below 88%, a low coverage of antiviral prophylaxis can lead to a slightly smaller number of hospitalizations, whereas for larger fitness values, the expected number of hospitalizations grows with growing prophylaxis coverage. In the case of 100% fitness, the expected number of hospitalizations grows from 314 per 100,000 (without prophylaxis) to 516 (20% receive prophylaxis). Apart from hoping that the population may benefit from antiviral prophylaxis via herd effects, the main goal of giving prophylaxis may be to maintain the work force of first responders. The expected work loss per person can approximately be calculated as follows: omitting antiviral treatment and prophylaxis, we expect about 75% of the population to become infected. One third of the infected individuals becomes severely sick for an average of 7 days and needs a further 5 days to recover. In total, 25% of all working adults should, therefore, be out of work for 12 days, i.e. we expect an average sickness related work loss of 3 days per person in this scenario. This value is further reduced by therapeutic antiviral treatment, partly because treated individuals recover more quickly, but more importantly because treatment prevents the transmission of infection and thereby has a beneficial herd effect. If all severe cases are treated, the expected work loss per person drops to 2.1 days if no NI resistance develops or if the NI resistant strain has a low fitness (with de novo development of a highly transmissible resistant strain, the expected work loss is only reduced to 2.4 days). As a reference value, the expected work loss per person without prophylaxis is depicted as grey horizontal bar in Fig. 3c. Values below the grey bar indicate that people who receive prophylaxis have a benefit, values above indicate that their work loss is higher than what would have been expected if nobody had received prophylaxis. If only few people receive prophylaxis, their work loss can be less than half of the value without prophylaxis, especially if the fitness of the resistant strain is low. For strains with a fitness of less than 92%, up to 20% may benefit from prophylaxis. For strains with higher fitness, the expected work loss can be worse than without prophylaxis if 10% or more receive prophylaxis.
Uncertainty analyses
In the following, we investigate the influence of parameter uncertainty by systematically varying the values of two unknown parameters:
(a) The susceptibility to infection x sus of people who take prophylaxis is varied from 0 to 100% (baseline value 50%). The results only deviate from the values shown in Figs. 3a–b by up to 1240 treatment failures and by up to 21 hospitalizations per 100,000 inhabitants (with a slight tendency towards higher deviations for combinations of high prophylaxis coverage and high fitness values). The parameter x sus strongly influences the expected work loss of people who take prophylaxis. If only few people take prophylaxis (cf. Fig. 3c), the expected work loss varies between nearly 0 days (x sus = 0) and 1.2 days (x sus = 1) for low fitness values, and from 0.9 to 1.6 days for high fitness values. If 10% of the population between 20 and 60 years of age take prophylaxis, the expected work loss varies from 0.1 day (x sus = 0) and 1.1 days (x sus = 1) for low fitness values, and it is about 2.2 days (irrespective of x sus ) for high fitness values.
(b) The fraction x imm of asymptomatic cases who develop protective immunity if they have received prophylaxis, is varied from 0 to 100% (baseline value 50%). The results only deviate from the values shown in Figs. 3a–c by up to 240 treatment failures, by up to 7 hospitalizations and by up to 0.1 days of work loss per person.
Quick calculation formula of the critical fitness of the resistant strain
The basic reproduction number R 0 is the expected number of people who are infected by a single index case in a fully susceptible population where no interventions are taken. If part of the population is immune and if interventions are taken, we talk about the effective reproduction number: the effective reproduction number of the drug sensitive strain and that of the NI resistant strain () can be approximated as follows:
Here, s is the susceptible fraction of the population, r SD is the reduction of transmission due to social distancing, r iso is the reduction due to case isolation, r tr is the reduction of transmission due to antiviral treatment of cases, r pro is the fraction of the population protected by prophylaxis and f is the fitness of the resistant strain (for more details, see the additional file SupplementaryData.doc). An infection can only be transmitted if the effective reproduction number is larger than 1. For the NI resistant strain, is larger than 1 if the fitness f is larger than (R 0 s (1- r SD )(1 - r iso ))-1. In a fully susceptible population (s = 1) where no interventions are performed (r SD = r iso = 0), the fitness f must be larger than = 40%. Immunity in the population and contact reductions increase the critical value of f. NI resistance only becomes a problem if the resistant strain spreads more efficiently than the drug sensitive one, i.e. if . This is the case if f > (1 - r tr )(1 - r pro ). Note that the critical value of f is independent of contact reduction measures; it only depends on the effects of treatment and prophylaxis. If the overall treatment effect is r tr = 0.188 (see additional file SupplementaryData.doc), the critical fitness of the resistant strain is about 81% if no prophylaxis is given. If the fitness exceeds this value, the resistant strain will invariably take over. The critical value for the fitness drops to 73% and 65%, respectively, if 10% or 20% of the population receive prophylaxis.
Discussion
Although NI resistant viruses may emerge de novo in treated patients in such a late state of their course of disease that most patients may not pass on the infection, our simulation study shows that the resistant strain will become highly prevalent in the population if its relative fitness is high and if NI treatment or prophylaxis are common. Prior to the 2007/8 influenza season, NI resistant strains were only infrequently found in patients after treatment with oseltamivir and in patients not exposed to oseltamivir. Early surveillance data from the 2007/8 influenza season on the northern hemisphere suggest that an oseltamivir resistant influenza virus type A (H1N1) circulates in several European countries and in North America [6, 7, 11]. Resistance infections have been reported from over two third of the countries which have implemented an influenza surveillance system and test for antiviral resistance. Furthermore, the proportion of resistant infections has become alarmingly high (between 4% and 70%) in the afflicted countries. Even in a fully susceptible population, a resistant virus can only spread if its fitness exceeded 40%. The populations in which the resistant seasonal influenza virus is spreading are far from being susceptible which further increases the minimum fitness of the resistant strain.
Considering the growing prevalence of resistant infections in spite of extremely low treatment rates, the fitness of the current resistant strain must indeed be very high. It is conceivable that a pandemic influenza strain may also become resistant without a considerable loss of transmissibility although this may be regarded as a worst case scenario. In a pandemic scenario, NI treatment is one of the major means of intervention and will be used extensively. Containing a potential pandemic within the country of its origin by widespread antiviral prophylaxis has been suggested [1, 3]. Prophylaxis has also been considered for local interventions after the international spread of the pandemic virus.
Even in the optimistic scenario where a drug sensitive infection is introduced into a population, de novo development of NI resistance in treated patients and the ensuing spread of resistant infections may lead to an early predominance of a resistant strain (Figs. 2a–c). De novo development of resistance in a person is a stochastic event and would demand for stochastic simulation in order to realistically describe the variability in the timing of such an event. Deterministic models like InfluSim only represent the average course the development of resistance in a population. For sake of simplicity, we have assumed that the development of resistance occurs in one step whereas other authors [18] assume that the first mutation leads to a resistant virus with impaired fitness and that the transmission fitness will gradually improve over time. Once a resistant virus with high fitness spreads in the population, the relatively rare de novo development of resistance in other people can be completely neglected because it is out-weighted by the multiplication of the virus in the population. This is because therapeutic and prophylactic NI use put pressure on the drug sensitive strain and favor the spread of circulating NI resistant infections. Figure 4 demonstrates this by comparing two scenarios: In scenario (a), NI resistance develops de novo by treatment of cases whereas in scenario (b) a NI resistant infection is introduced 4 weeks later, but no de novo development of NI resistance occurs. The resulting curves are nearly indistinguishable, indicating that once a resistant strain of high transmissibility spreads in a population where there is a lot of pressure on drug sensitive infection, any further de novo development of resistance can be neglected. Our calculations show that under widespread treatment, the NI resistant strain spreads faster than the non-resistant one if its fitness exceeds 81%. Prophylaxis will further increase the pressure, leading to a quicker replacement of the drug sensitive strain by the resistant one and increasing the number of unsuccessfully treated patients (Fig. 3a). If the fitness of the resistant virus is between 90 and 100%, prophylaxis even increases the total number of cases (Figs. 2a–c) and hospitalizations (Fig. 3b), and we obtain the counter-intuitive result that the work loss of those people who receive prophylaxis may become larger than without prophylaxis (Fig. 3c). Our simulations assume that a small fraction of the population receives prophylaxis during the whole course of the epidemic whereas individuals are only advised to take prophylaxis for a maximum of six weeks [19]. Another approach would be to split up the group of first responders into subgroups who alternatively receive prophylaxis, but the pressure on the drug sensitive virus exerted by prophylaxis mainly depends on the prophylaxis coverage and does not change much if different people receive the drug at different times. Especially for low prophylaxis coverage, our results should be relatively robust in spite of this over-simplification. Our simulation results confirm other authors' findings which indicate that the benefits of antiviral drug use to control pandemic influenza may be reduced by NI resistance in the virus [5, 6]. Our model structure, assumptions and parameter values differ from those reports: we used higher values for R 0 and for de novo development of NI resistance than [5], but lower values than [6], and we used a wide range of prophylaxis levels. In contrast to [5] and [6], we found clear detrimental effects of any level of prophylaxis if the relative fitness of the resistant strain is higher than 80% [20, 21].
Conclusion
Surveillance of the appearance of NI resistant infections and, most importantly, of the fitness of resistant strains will be crucial to manage a pandemic wave. Uncontrolled use of NI may do more harm than good. If a NI resistant pandemic strain is reported to spread, the use of NI should mainly be restricted to the treatment of cases whereas prophylaxis must be reduced to an absolute minimum. If the fitness of the NI resistant pandemic strain is high, any use of prophylaxis may increase the number of hospitalizations and deaths in the population. Thus, public health systems should diversify intervention strategies by using prophylaxis in a very restrictive manner, by stockpiling different types of antiviral drugs and by supplementing pharmaceutical interventions with social distancing measures [22, 23].
References
Ferguson NM, Cummings DA, Cauchemez S, Fraser C, Riley S, Meeyai A, Iamsirithaworn S, Burke DS: Strategies for containing an emerging influenza pandemic in Southeast Asia. Nature. 2005, 437 (7056): 209-214. 10.1038/nature04017.
Longini IM, Halloran ME, Nizam A, Yang Y: Containing pandemic influenza with antiviral agents. Am J Epidemiol. 2004, 159 (7): 623-633. 10.1093/aje/kwh092.
Longini IM, Nizam A, Xu S, Ungchusak K, Hanshaoworakul W, Cummings DA, Halloran ME: Containing pandemic influenza at the source. Science. 2005, 309 (5737): 1083-1087. 10.1126/science.1115717.
Alexander ME, Bowman CS, Feng Z, Gardam M, Moghadas SM, Rost G, Wu J, Yan P: Emergence of drug resistance: implications for antiviral control of pandemic influenza. Proc R Soc Lond B Biol Sci. 2007, 274 (1619): 1675-1684. 10.1098/rspb.2007.0422.
Lipsitch M, Cohen T, Murray M, Levin BR: Antiviral Resistance and the Control of Pandemic Influenza. PLoS Med. 2007, 4 (1): e15-10.1371/journal.pmed.0040015.
Regoes RR, Bonhoeffer S: Emergence of drug-resistant influenza virus: population dynamical considerations. Science. 2006, 312 (5772): 389-391. 10.1126/science.1122947.
Stilianakis NI, Perelson AS, Hayden FG: Emergence of drug resistance during an influenza epidemic: insights from a mathematical model. J Infect Dis. 1998, 177 (4): 863-873.
Blower SM, Gershengorn HB, Grant RM: A tale of two futures: HIV and antiretroviral therapy in San Francisco. Science. 2000, 287 (5453): 650-654. 10.1126/science.287.5453.650.
Blower SM, Porco TC, Darby G: Predicting and preventing the emergence of antiviral drug resistance in HSV-2. Nat Med. 1998, 4 (6): 673-678. 10.1038/nm0698-673.
Yen HL, Ilyushina NA, Salomon R, Hoffmann E, Webster RG, Govorkova EA: Neuraminidase inhibitor-resistant recombinant A/Vietnam/1203/04 (H5N1) influenza viruses retain their replication efficiency and pathogenicity in vitro and in vivo. J Virol. 2007, 81 (22): 12418-12426. 10.1128/JVI.01067-07.
Lackenby A, Hungnes O, Dudman SG, Meijer A, Paget WJ, Hay AJ, Zambon MC: Emergence of resistance to oseltamivir among influenza A(H1N1) viruses in Europe. Euro surveill. 2008, 13 (5):
Eichner M, Schwehm M, Duerr HP, Brockmann SO: The influenza pandemic preparedness planning tool InfluSim. BMC Infect Dis. 2007, 7: 17-10.1186/1471-2334-7-17.
InfluSim. [http://www.influsim.info]
Chowell G, Nishiura H, Bettencourt LM: Comparative estimation of the reproduction number for pandemic influenza from daily case notification data. J R Soc Interface. 2007, 4 (12): 155-166. 10.1098/rsif.2006.0161.
Kaiser L, Wat C, Mills T, Mahoney P, Ward P, Hayden F: Impact of oseltamivir treatment on influenza-related lower respiratory tract complications and hospitalizations. Arch Intern Med. 2003, 163 (14): 1667-1672. 10.1001/archinte.163.14.1667.
Aoki FY, Boivin G, Roberts N: Influenza virus susceptibility and resistance to oseltamivir. Antivir Ther. 2007, 12 (4 Pt B): 603-616.
Tamiflu. Arzneimittel-Kompendium der Schweiz. [http://www.kompendium.ch]
Handel A, Regoes RR, Antia R: The role of compensatory mutations in the emergence of drug resistance. PLoS Comput Biol. 2006, 2 (10): e137-10.1371/journal.pcbi.0020137.
Updated review of influenza antiviral medicinal products for potential use during pandemic by the Committee for Medicinal Products for Human Use (CHMP) of the European Medicines Agency (EMEA). [http://www.emea.europa.eu/htms/human/pandemicinfluenza/antivirals.htm]
Moghadas SM, Bowman CS, Rost G, Wu J: Population-wide emergence of antiviral resistance during pandemic influenza. PLoS ONE. 2008, 3 (3): e1839-10.1371/journal.pone.0001839.
Moghadas SM: Management of drug resistance in the population: influenza as a case study. Proc R Soc B. 2008
Duerr HP, Brockmann SO, Piechotowski I, Schwehm M, Eichner M: Influenza pandemic intervention planning using InfluSim: pharmaceutical and non-pharmaceutical interventions. BMC Infect Dis. 2007, 7 (1): 76-10.1186/1471-2334-7-76.
McCaw JM, Wood JG, McCaw CT, McVernon J: Impact of emerging antiviral drug resistance on influenza containment and spread: influence of subclinical infection and strategic use of a stockpile containing one or two drugs. PLoS ONE. 2008, 3 (6): e2362-10.1371/journal.pone.0002362.
Pre-publication history
The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-2334/9/4/prepub
Acknowledgements
This work has been supported by a project of the SFOPH (contract no. 06.001333/304.0001-108), and by the EU projects SARScontrol (FP6 STREP; contract no. 003824) (HPD) and INFTRANS (FP6 STREP; contract no. 513715) (MS). We thank M. Mäusezahl and H. C. Matter for their support and for reviewing a previous version of the manuscript.
Author information
Authors and Affiliations
Corresponding author
Additional information
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
ME was responsible for model design, mathematical formulas and the writing of manuscript. MS designed and programmed the software and provided the simulation results. HPD advised on modeling questions, helped preparing the figures and proofread the manuscript. MW gave epidemiologic advice and co-wrote the manuscript. DK and SOB defined the research questions, made public health decisions and proofread the manuscript. BV coordinated the research project and co-wrote the manuscript. All authors have read the manuscript and have given final approval of the current version to be published.
Electronic supplementary material
Authors’ original submitted files for images
Below are the links to the authors’ original submitted files for images.
Rights and permissions
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
About this article
Cite this article
Eichner, M., Schwehm, M., Duerr, HP. et al. Antiviral prophylaxis during pandemic influenza may increase drug resistance. BMC Infect Dis 9, 4 (2009). https://doi.org/10.1186/1471-2334-9-4
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/1471-2334-9-4