Site Loader
Rock Street, San Francisco

Seroepidemiological studies aim to understand population-level exposure and immunity to infectious diseases. Their results are normally presented as binary outcomes describing the presence or absence of pathogen-specifc antibody, despite the fact that many assays measure continuous quantities.

A population’s natural distribution of antibody titers to an endemic infectious disease may include information on multiple serological states – naiveté, recent infection, non-recent infection, childhood infection – depending on the disease in question and the acquisition and waning patterns of immunity. In this study, we investigate 20,152 general-population serum samples from southern Vietnam collected between 2009 and 2013 from which we report antibody titers to the infuenza virus HA1 protein using a continuous titer measurement from a protein microarray assay. We describe the distributions of antibody titers to subtypes 2009 H1N1 and H3N2. Using a model selection approach to ft mixture distributions, we show that 2009 H1N1 antibody titers fall into four titer subgroups and that H3N2 titers fall into three subgroups. For H1N1, our interpretation is that the two highest-titer subgroups correspond to recent and historical infection, which is consistent with 2009 pandemic attack rates. Similar interpretations are available for H3N2, but right-censoring of titers makes these interpretations difcult to validate.Te distribution of antibodies in a human population is a fossil imprint of the population’s past exposure to infectious disease. If individuals’ antibody concentrations can be measured accurately, they can be used to infer both the size and timing of past epidemics.

Best services for writing your paper according to Trustpilot

Premium Partner
From $18.00 per page
4,8 / 5
Writers Experience
Recommended Service
From $13.90 per page
4,6 / 5
Writers Experience
From $20.00 per page
4,5 / 5
Writers Experience
* All Partners were chosen among 50+ writing services by our Customer Satisfaction Team

Te two key post-epidemic processes that need to be measured to make this inference possible are the rate of antibody acquisition and the rate of antibody waning. Te rate of antibody acquisition post-infection is rapid (weeks) for most viral pathogens, but more difcult to measure for more complex pathogens that present the immune system with a diverse set of antigens. Te rate of antibody waning, however, is rarely measured even for viral pathogens. To correctly translate a population’s antibody titer distribution to its epidemic history, accurate measures of both these rates are necessary. To validate that this reconstruction has been done correctly, a large cohort with long-term follow-up and precise antibody measurements would be required. Studies like these are difcult to run and difcult to fnd in the scientifc literature – both in methodological development and feld implementation.

Further complicating the issue is that antibody measurements are rarely 100% specifc, and that low-level cross-reactive antibodies ofen are ignored by setting a cut-of for positivity.To begin investigating what an antibody distribution can tell us about a population’s epidemic history, we initiated a large-scale time-structured serological survey1, 2 and an observational clinical study that includes repeat patient follow-ups to measure rates of antibody waning3 ; the results of the serological survey are presented here. Infuenza A virus was chosen as the pathogen of interest as (i) it is an important, globally-circulating human pathogen, (ii) infuenza is well characterized antigenically, (iii) a precise and repeatable serological assay was available, and (iv) the human population receives almost no infuenza vaccination in our study location of southern Vietnam. Te frst aim of this study was to move away from the binary approach to serology – which classifes individuals as seropositive or seronegative4–8 – and to describe the underlying structure of a general-population antibody-titer distribution by assuming that an individual can belong to any number of serological states.The rationale for a detailed descriptive analysis of antibody titer distributions is that titer groups or titer ranges may be able to provide diferentiating information on the type of infection, e.g.

recent versus non-recent infections, or primary versus non-primary. Te binary approach of classifying individuals as seropositive and seronegative is not as informative as it could be given the richness of some serological datasets, and it is already known to have two practical drawbacks. First, the cutof value for seropositivity is typically calibrated from a group of patients with confrmed acute infection, by collecting convalescent serum samples a few weeks or a few months afer symptoms onset. Tis means that the correct application of the cut-of value is the identifcation of recent symptomatic infections rather than any past infections. Tus, applying this threshold to a population-wide serological cross-section will likely result in an underestimate of the seroprevalence. Second, binary classifcation in serology results in incorrect or inconclusive classifcations for samples with borderline measurements8–10. Non-binary analyses of serological data are present in the literature for a range of pathogens10–18 including infuenza virus19, 20, but very few of these studies are able to look at non-vaccinated populations and none have the scale and precision presented here.In the present study, we analyze a large set of general-population serum samples collected as residual serum from biochemistry and haemotology labs in four hospitals in southern Vietnam, from 2009 to 2013.

Using a zero-infated mixture modeling approach, we allow for up to seven serological states. To account for the large sample size in our model selection procedure, we use the Bayesian Information Criterion, and to avoid inference of spurious serological states we set additional criteria to ensure that inferred titer groups are epidemiologically meaningful. We hypothesized that serological classifcation of infuenza antibody titers would be non-binary and that age and lineage exposure (H1N1 only) would be associated with certain titer groups.

We found that H1N1 antibody titer distribution are best classifed into four titer groups, that H3N2 is best classifed into three groups, and that censoring may have prevented a complete classifcation of H3N2 titers.


A total of 20,152 sera were collected and tested for antibody concentrations by protein microarray. Te samples represent patients attending hospitals in four cities – Ho Chi Minh City (n=5788), Nha Trang (n=5630), Buon Ma Tuot (n=4144), and Hue (n=4590) in central and southern Vietnam.

Titer distributions varied by age, as expected (Fig. 1) but did not vary by site (Figures S8 and S9). Figure 1 shows the age-stratifed titer distributions to the HA1 component of the 2009 H1N1 virus and the most recently circulating H3N2 variants.

If individuals truly represented seropositive (exposed) and seronegative (unexposed or naïve) categories, a mixture model of two components would classify samples into two subgroups. Visually, this does not appear to be the case as a broad range of titers was observed for both subtypes across all age groups. Tus, a mixture distribution ftting approach was employed to determine the appropriate number of components necessary to accurately describe the titer data.Mixture distribution fts for up to six components, with an additional weight at a log-titer of one (“zero infation”), are shown in Fig.

 2 for H1N1 and Fig. S10 for H3N2. For both subtypes, it is clear that a binary classifcation of titer is not the most informative interpretation of the titer distribution, as both the one- and two-component models (top two rows) did not capture the underlying structure of the dataset adequately.

When stratifying the data by site (sample size >4,000), the Bayesian Information Criterion (BIC) selected four components as the best model for the H1N1 data (fve for Hue, but the ?BIC=18 here was relatively small compared to other changes between nested models) and three components as the best model for H3N2. Te fve- and six-component models either overft the data (according to the BIC) or included low-variance/low-weight components, which would correspond to an implausible population subgroup with a very specifc antibody titer (Figs. 2 and 3). Tis was readily seen in the aggregate data which is why the BIC-selected models of the by-site data are likely to be better explanations of the structure of these titer distributions. BIC improvement from n mixture components to n+1 components is shown in Table 1 for 2009 H1N1 and Table S7 for H3N2. Te means and variances were allowed to be free in these analyses, and the confdence intervals for the inferred parameters (Appendix Section 7) suggest that the structure of the distributions and the inferred values were robust across the four sites in our analysis.

The three- and four-component mixtures indicate that these data can be used to develop a more informative serological classifcation for infuenza. Using known results for this microarray assay3, 20, 21, titers below 100 would be classifed as negative or ‘not previously exposed to this particular infuenza strain’. For H1N1, this indicates that titers in the frst component ?1=29.8 (95% CI 29.1–30.5) and in the second component ?2=75.0 (95% CI 73.

4–76.7) would both correspond to seronegative individuals. Similarly, for H3N2, seronegative individuals would be represented by the frst component ?1=80.2 (95% CI 76.

7–83.4). Te second-highest titer component has mean ?3=247.3 (95% CI 240.

8–261.7) for H1N1 and ?2=213.3 (95% CI 209.

7–216.6) for H3N2. Te highest titer component has mean ?4=670.9 (95% CI 519.8–787.

9) for H1N1 and ?3=455.0 (95% CI 428.1–483.7) for H3N2. Te natural interpretation of these high-titer subgroups – based on antibody titers measured as a function of time since infection3 – is that they represent more recent infections.

As it is known that the infuenza antibody decay rate is fast enough to be observed in the frst six to twelve months afer an acute infection22, 23, for H1N1 the highest titer subgroup may be an approximate designation for recently infected individuals, and the second highest titer subgroup may correspond to ‘historically infected’ individuals, i.e. individuals infected at some point in the non-recent past.Visualization of model selection process for 2009 H1N1 titer-distribution models from Fig. 2. Te y-axes show the ftted values of wi (mixture weights), ?i (means), and ?i (standard deviations).</p>
<p> Components’ shades were ranked from lightest to darkest in the order of increasing ?. In the top panel, the “0th component” represents the point mass w0 placed at 20 for titers below the lower detection limit of 20. Note that in many cases for fve or six components, the weights or standard deviation parameters are close to zero; for some cases, two of the inferred mean parameters are very close to each other.” width=”623″ height=”323″ /><strong>Figure 3.</strong> Visualization of model selection process for 2009 H1N1 titer-distribution models from Fig. 2.</p>
<p> Te y-axes show the ftted values of wi (mixture weights), ?i (means), and ?i (standard deviations). Components’ shades were ranked from lightest to darkest in the order of increasing ?. In the top panel, the “0th component” represents the point mass w0 placed at 20 for titers below the lower detection limit of 20. Note that in many cases for fve or six components, the weights or standard deviation parameters are close to zero; for some cases, two of the inferred mean parameters are very close to each other.<img class=Discussion

Using a large collection of serum samples and a continuous measurement of antibody titer, we were able to describe the natural distribution of antibody titers to the 2009 H1N1 and H3N2 subtypes of infuenza virus. As there is almost no infuenza vaccination in Vietnam and as infuenza in Vietnam is characterized by a combination of local persistence and annual/biannual outbreaks26–28, characterization of titer distribution in this context is a useful general approach for looking at the immune status of a population at quasi-equilibrium with an endemic infectious disease. With a mixture model approach, we were able to identify the presence of multiple exposure groups in the population according to their titers. Our interpretation of these multiple exposure groups – according to titers measured for confrmed cases3, 21 and past measurements of the rate of antibody waning22, 23 – is that they represent recently infected individuals, historically (i.e. non-recently) infected individuals, and naïve individuals. Note that for infuenza, a naïve individual is one who has not been exposed to the currently circulating strain, which means that there will be naïve individuals in all age groups.

This study used an atypical seroepidemiological design as the samples were collected continuously, and not specifcally in a post-epidemic or post-pandemic scenario. In addition, the serum samples were collected in the tropics where continuous circulation of infuenza virus is believed to occur28–33 and where populations are much less likely to be vaccinated for infuenza (less than 0.8% annual coverage for Vietnam).

Terefore, the present data set is the frst to show the natural distribution of infuenza antibody titers in a human population.One useful application of these results in future serological studies is to encourage, by default, the inclusion of multiple serological states in the data analysis phase, which may result in a more informative classifcation of antibody titer than a separation into seropositive and seronegative. Te classifcation proposed here uses antibody levels as proxies for recency of infection, and if correct, this should allow for a more informative reconstruction of the population’s epidemic history.

In general, knowing the IgG antibody waning rate is essential for interpreting the titers measured in serological cross-sections34, 35, and using waning rates to estimate the time of past infection has already been attempted for some infectious diseases36–39, but not for infuenza virus. Longitudinal follow-up studies that are able to provide accurate estimates of antibody waning rates are crucial for this type of analysis, but they are rare3, 23, 40, 41.Two major limitations of serological classifcation systems will need to be better understood. First, a mixture distribution approach does not guarantee that individuals can be easily classifed into one of several titer subgroups. With substantial overlap in some mixture components, individuals can have approximately equal probabilities of belonging to two or three diferent titer categories. In addition, individual variation will have a large efect on titer interpretations. A high-titer sample could represent a recent infection, but individuals can maintain high titers longer than the mean duration observed in clinical studies. Tis would normally, but not exclusively, be observed in children.

Likewise, lower antibody titers (in the 200–250 range) could indicate historical past infection, a low response to a recent infection42, or a recent but mild infection. With serological data alone, these scenarios cannot be distinguished. For subtype H3N2 specifcally, low titer levels could indicate cross-reactions between antibodies generated to an older infuenza variant than the recent H3N2 HA1 proteins spotted on the protein microarray.Second, a major challenge in infuenza seroepidemiology is that it is difcult to take into account the efects of original antigen sin42, 43 or age-dependent seroconversion40 (ADS). Age-dependent seroconversion is distinct from original antigenic sin in that ADS assumes that individuals of diferent ages seroconvert to diferent titer levels irrespective of the individual’s infection history. In principle, the efect of ADS should be detectable for 2009 H1N1 infections in individuals younger than 50, as for these individuals an exposure to the 2009 virus would have been a frst exposure.

However, the mixture component means (?i parameters) and the component weights (wi ) are not separately identifable in the mixture model. Tus, we cannot state that the ‘recently infected’ titer subgroups are comparable across age groups, as the inferential process will make the exact defnition of recency diferent for the 10–19 age group than for the 20–44 age group. Even if we were to assume that the fourth mixture components should be comparable across age groups, the titer means denoted by ?4 in Fig. 4 do difer but are within one standard deviation of one another. Tus, there is a lack of evidence for ADS in our titer data. As we only considered recent antigens in this analysis, efects of original antigenic sin were not able to be investigated.

Titer histograms and ft results for mixture models with diferent numbers of components (label on the lef is the number of mixture components) and grouped by diferent age groups recommended by the CONCISE (http://consise.</p>
<p> consortium for 2009 H1N1 infuenza. Histograms are weighted to adjust for age and gender according to the Vietnam national housing census in 2009.</p>
<p> Te numbers in the upper right corner of each panel are the ftted BIC scores of the respective model. For each panel, the blue lines are the normalized probability density of the component distributions with darker colors used for increasing ?. Black lines are the total mixture distribution density; and the black dots are estimated probability weight of the mixture model for titers ?7.0. Te fractions of individuals with titers below the detection limit of 20 and above 1280 that were out of the plotting ranges are shown next to their respective bars.” width=”632″ height=”318″ /><strong>Figure 4.</p>
<p></strong> Titer histograms and ft results for mixture models with diferent numbers of components (label on the lef is the number of mixture components) and grouped by diferent age groups recommended by the CONCISE ( consortium for 2009 H1N1 infuenza. Histograms are weighted to adjust for age and gender according to the Vietnam national housing census in 2009. Te numbers in the upper right corner of each panel are the ftted BIC scores of the respective model. For each panel, the blue lines are the normalized probability density of the component distributions with darker colors used for increasing ?. Black lines are the total mixture distribution density; and the black dots are estimated probability weight of the mixture model for titers ?7.</p>
<p>0. Te fractions of individuals with titers below the detection limit of 20 and above 1280 that were out of the plotting ranges are shown next to their respective bars.Te next critical step in this analysis will be using titer data from follow-up on confrmed cases3 to determine if the natural distribution of antibody titers conforms to the recent, historical, and naïve categories as presented here. If antibody waning rates can be measured with a high degree of precision, these may allow for a detailed description of individuals’ recency of infection and possibly a reconstruction of past epidemic history in human populations. Large-scale serological studies like the one presented here are labor-intensive and slow to generate results. Nevertheless, the long follow-up and the large sample size will be worth it if seroepidemiology can be pushed forward to maximize the amount of biological information that can be extracted from population-level serology studies.</p>
<h3 style=Materials and Methods

Residual serum samples were collected from four hospital laboratories in southern Vietnam: the Hospital for Tropical Diseases in Ho Chi Minh City (urban, densely populated), Khanh Hoa Provincial Hospital in Nha Trang city (small urban, central coast), Dak Lak Provincial Hospital in Buon Ma Tuot city (central highlands, rural), and Hue Central hospital in Hue City (small urban, central coast). Samples were collected from July 2009 to December 2013 on a bimonthly basis; 200 were included in each collection from all age groups (neonates to elderly individuals in their 90s).

Samples were anonymized, delinked, and labeled with age, gender, originating hospital ward (HIV wards were excluded), and date of collection. Samples were collected from both inpatients and outpatients and are believed to represent the hospital-going population in their respective cities.This assumption is currently being tested and will continue to be tested as diferent antibody assays are performed on the sample set. Two early analyses (one unpublished and one published44) suggest that when looking at hospital presentation with hepatitis, the younger age range (<20) in the sample set may represent a sub-population more vulnerable to infectious disease exposure than the general population. Te sample collection described here is part of a large ongoing study in serial seroepidemiology1, 2, 34 aimed at describing the dynamics of infuenza circulation in southern Vietnam. The study was approved by the Scientifc and Ethical Committee of the Hospital for Tropical Diseases in Ho Chi Minh City and the Oxford Tropical Research Ethics Committee at the University of Oxford.The samples were tested for presence of infuenza antibodies using a protein-microarray (PA) method45, at serial four-fold dilutions from 20 to 1280, to test for IgG antibody to the HA1 component of 16 diferent infuenza viruses1 .

Two-fold dilutions were used in some instances; see validation of this approach in Appendix Section 2. A sample of the international standard (IS) for testing antibody response to infuenza A H1N1 Pandemic 2009 (H1-09) was included on every slide to correct for inter-laboratory, inter-technician, and inter-slide variations45 (Appendix Section 1.2). Assay repeatability was assessed using a positive control and replicates of patient samples (Appendix Section 3). Titers were defned as the dilution at which samples yield a median response between the minimum and maximum luminescence values of 3000 and 65535.

Titers of all human samples on each slide are normalized based on the IS titers of the reference antigen against its geometric mean (Table S2). In this analysis, titers to the 2009 H1N1 virus (A/California/6/2009) and recently circulating H3N2 viruses (geometric mean titer to A/Victoria/210/2009 and A/Victoria/361/2011) were analyzed.To describe the distribution of infuenza antibody titers in the Vietnamese population, titer values were separated by site, adjusted to their province’s age and gender distribution46 (Appendix Section 4), and plotted as a simple weighted histogram (Fig. 1). A series of mixture models was used to ft this distribution, with the assumption being that individual samples have one of several immune statuses which are represented by the diferent components in the mixture model. Our hypothesis was that the sample population consists of diferent subpopulations with diferent antibody levels depending on their infection history and that each of these components can be represented by a single parametric distribution.Titers were log-transformed and assumed to come from a C-component mixture distribution with the corresponding likelihood:in which the si parameters are sampling corrections to adjust the sample age and sex distribution to the population’s true demographic distribution; fj (xi |?j ), j=1, 2, .., C is the probability density function that a given sample xi belongs to the jth-component in the mixture. C is the number of mixture components47, 48.

The microarray assay produces continuous log-titer results between 1.0 (titer of 20) and 7.0 (titer of 1280). To account for these detection limits, an extra probability weight w0 was added at 20 to account for samples that had antibody concentrations at or below the detection limit of 20.

Tis can be considered a zero-infated mixture model, where titers of 20 are the “zeroes”. Because of this added probability mass, we discretized the probability mass functions to make the entire distribution discrete; hence the distributions f formally represents discretized versions of continuous density functions (Appendix Section 5). At the upper detection limit of 7.0, the mixture distribution was censored assuming that individuals with titers of 7.0 represented a class of seropositive individuals with a real titer value if the assays had been continued to be diluted until the real titer was found. Censoring on the right and truncating on the lef gave the best ft (according to BIC) among the four combinations. Truncating on the lef means that the extra weight on the lef-hand side of the probability density function (the portion below 20) was simply discarded when performing the fts, as “zero-infation” on the lef-hand side was used to ft the number of samples that had titers of 20 or below.Thus, the log-likelihood in (2) was modifed as:References

  1. Boni, M. F.

    et al. Population-level antibody estimates to novel infuenza A/H7N9. J Infect Dis 208, 554–558 (2013).

  2. Todd, S., de Bruin, E., Nhat, N.

    T. D., Koopmans, M. & Boni, M. F. Reply to Pawar et al.–Immunity Status Against Infuenza A Subtype H7N9 and Other Avian Infuenza Viruses in a High-Risk Group and the General Population in India. J Infect Dis 210, 161–163 (2014).

  3. Todd, S. Infuenza in Vietnam: cross immunity from prior infection and its efect on population-level estimates of infection. (University of Liverpool and Liverpool School of Tropical Medicine, PhD Tesis, 2016).
  4. Steens, A. et al. Age-dependent patterns of infection and severity explaining the low impact of 2009 infuenza A (H1N1): Evidence from serial serologic surveys in the Netherlands. Am J Epid 174, 1307–1315 (2011).

  5. Chen, M. I. C. et al. 2009 Infuenza A(H1N1) Seroconversion Rates and Risk Factors Among Distinct Adult Cohorts in Singapore. J Am Med Assoc 303, 1383–1391 (2010).
  6. Wu, J. T.

    et al. Te Infection Attack Rate and Severity of 2009 Pandemic H1N1 Infuenza in Hong Kong. Clin Infect Dis 51, 1184–1191 (2010).

  7. Miller, E.

    et al. Incidence of 2009 pandemic infuenza A H1N1 infection in England: a cross-sectional serological study. Lancet 375, 1100–1108 (2010).

  8. Cauchemez, S.

    et al. Infuenza Infection Rates, Measurement Errors and the Interpretation of Paired Serology. PLoS Pathog 8, e1003061 (2012).

  9. Broberg, E.

    , Nicoll, A. & Amato-Gauci, A. Seroprevalence to infuenza A(H1N1) 2009 virus – Where are we? Clin Vacc Immunol 18, 1205–1212 (2011).

  10. Baughman, A. L., Bisgard, K.

    M., Lynn, F. & Meade, B. D. Mixture model analysis for establishing a diagnostic cut-of point for pertussis antibody levels. Stat.

    Med. 25, 2994–3010 (2006).

  11. Gay, N. J., Vyse, A. J., Enquselassie, F.

    , Nigatu, W. & Nokes, D. J.

    Improving sensitivity of oral fuid testing in IgG prevalence studies: application of mixture models to a rubella antibody survey. Epidemiol Infect 285–291 (2003).

  12. Neuenschwander, B. E., Zwahlen, M., Kim, S. J., Lee, E.

    G. & Rieder, H. L. Determination of the prevalence of infection with Mycobacterium tuberculosis among persons vaccinated against Bacillus Calmette-Guerin in South Korea.

    Am J Epidemiol 155, 654–663 (2002).

  13. Hardelid, P. et al. Analysis of rubella antibody distribution from newborn dried blood spots using fnite mixture models.

    Epidemiol Infect 136, 1698–1706 (2008).

  14. Greiner, M., Franke, C.

    R., Böhning, D. & Schlattmann, P. Construction of an intrinsic cut-of value for the sero-epidemiological study of Trypanosoma evansi infections in a canine population in Brazil: a new approach towards an unbiased estimation of prevalence.

    Acta Trop. 56, 97–109 (1994).

  15. Rota, M.

    C. et al. Measles serological survey in the Italian population: Interpretation of results using mixture model. Vaccine 26, 4403–4409 (2008).

  16. Vyse, A. J.

    , Gay, N. J., Hesketh, L. M.

    , Morgan-Capner, P. & Miller, E. Seroprevalence of antibody to varicella zoster virus in England and Wales in children and young adults. Epidemiol. Infect. 132, 1129–34 (2004).

  17. Fujii, Y.

    et al. Serological Surveillance Development for Tropical Infectious Diseases Using Simultaneous Microsphere-Based Multiplex Assays and Finite Mixture Models. PLoS Negl Trop Dis 8 (2014).

  18. Liu, Y. et al. Sero-epidemiology of measles in general population in Jiangsu province of China: Application of mixture models to interpret the results from a cross-sectional study.

    Vaccine 29, 1000–1004 (2011).

  19. te Beest, D. E., Birrell, P. J., Wallinga, J.

    , De Angelis, D. & van Boven, M. Joint modelling of serological and hospitalization data reveals that high levels of pre-existing immunity and school holidays shaped the infuenza A pandemic of 2009 in Te Netherlands. J R Soc Interface 12 (2015).

  20. te Beest, D. et al. Discrimination of infuenza infection (A/2009 H1N1) from prior exposure by antibody protein microarray analysis. PLoS One 9 (2014).

  21. Huijskens, E. G. W.

    et al. Profling of Humoral Response to Infuenza A (H1N1) pdm09 Infection and Vaccination Measured by a Protein Microarray in Persons with and without History of Seasonal Vaccination. PLoS One 8, e54890 (2013).

  22. Ng, S.

    et al. Estimation of the association between antibody titers and protection against confrmed infuenza virus infection in children. J Infect Dis 208, 1320–1324 (2013).

  23. Horsfall, F. L. Jr. & Rickard, E. R.

    Neutralizing antibodies in human serum afer infuenza A: Te lack of strain specifcity in the immunological response. J Exp Med 74, 433–439 (1941).

  24. Horby, P. et al. Te Epidemiology of Interpandemic and Pandemic Infuenza in Vietnam, 2007–2010: Te Ha Nam Household Cohort Study I. Am J Epidemiol 175, 1062–1074 (2012).
  25. Hien, T. T.

    et al. Early Pandemic Infuenza (2009 H1N1) in Ho Chi Minh City, Vietnam: A Clinical Virological and Epidemiological Analysis. PLoS Med 7, e1000277 (2010).

  26. Lam, H. M.

    et al. Non-annual seasonality of infuenza-like illness in a tropical urban setting. bioRxiv 100222 (2017).

  27. Nguyen, Y. T. et al. National surveillance for infuenza and infuenza-like illness in Vietnam, 2006?2010. Vaccine 31, 4368–4374 (2013).
  28. Le, M. Q.

    et al. Migration and Persistence of Human Infuenza A Viruses, Vietnam, 2001–2008. Emerg Infect Dis 19, 1756–1765 (2013).

  29. Viboud, C., Alonso, W. J.

    & Simonsen, L. Infuenza in tropical regions. PLoS Med 3, e89 (2006).

  30. Alonso, W.

    J. et al. Seasonality of infuenza in Brazil: a traveling wave from the Amazon to the subtropics. Am J Epidemiol 165, 1434–42 (2007).

  31. Lin, J. et al.

    Infuenza seasonality and predominant subtypes of infuenza virus in Guangdong, China, 2004–2012. J Torac. Dis 5, 2004–2012 (2013).

  32. Nelson, M. I. et al.

    Multi-year persistence of two pandemic A/H1N1 infuenza virus lineages in West Africa. J Infect Dis 201, 121–125 (2014).

  33. Cheng, X. et al.

    Epidemiological dynamics and phylogeography of infuenza virus in southern china. J Infect Dis 207, 106–14 (2013).

  34. Vinh, D. N. & Boni, M. F.

    Statistical identifability and sample size calculations for serial seroepidemiology. Epidemics 12, 30–39 (2015).

  35. Wu, J. T. et al. Inferring infuenza infection attack rate from seroprevalence data. PLoS Pathog 10, e1004054 (2014).

  36. Borremans, B., Hens, N., Beutels, P.

    , Leirs, H. & Reijniers, J. Estimating Time of Infection Using Prior Serological and Individual Information Can Greatly Improve Incidence Estimation of Human and Wildlife Infections. PLoS Comput Biol 12, 1–18 (2016).

  37. de Melker, H. E., Versteegh, F.

    G. A., Schellekens, J. F. P., Teunis, P. F.

    M. & Kretzschmar, M. Te incidence of Bordetella pertussis infections estimated in the population from a combination of serological surveys. J Infect 53, 106–113 (2006).

  38. Simonsen, J. et al. Estimation of incidences of infectious diseases based on antibody measurements. Stat Med 28, 1882–1895 (2009).
  39. Teunis, P. F. M. et al. Biomarker dynamics: Estimating infection rates from serological data. Stat Med 31, 2240–2248 (2012).
  40. Versteegh, F. G. A. et al. Age-specifc long-term course of IgG antibodies to pertussis toxin afer symptomatic infection with Bordetella pertussis. Epid Infect 133, 737–748 (2005).
  41. Strid, M. A. et al. Antibody Responses to CampylobacterInfections Determined by an Enzyme-Linked Immunosorbent Assay: 2-Year Follow-Up Study of 210 Patients Antibody Responses to Campylobacter Infections Determined by an Enzyme-Linked Immunosorbent Assay: 2-Year Follow-Up. Clin. Diagn. Lab. Immunol. 8, 314–319 (2001).
  42. Lessler, J. et al. Evidence for Antigenic Seniority in Infuenza A (H3N2) Antibody Responses in Southern China. PLoS Pathog 8, e1002802 (2012).
  43. Fazekas De St. Groth, S. & Webster, R. G. Disquisitions on Original Antigenic Sin. I. Evidence in Man. J Exp Med 124, 331–345 (1966).
  44. Berto, A. et al. Hepatitis E in southern Vietnam: seroepidemiology in humans and molecular epidemiology in pigs. Zoonoses Public Health (2017).
  45. Koopmans, M. et al. Profling of humoral immune responses to infuenza viruses by using protein microarray. Clin Microbiol Infect 18, 797–807 (2012).
  46. Government Statistics Ofce of Vietnam. Vietnam population and housing census 2009, age-sex structure and marital status of the population in Vietnam. (2011).
  47. Hens, N., Shkedy, Z., Aerts, M. & Faes, C. Modeling infectious disease paramters based on serological and social contact data. (Springer, 2012).
  48. McLachlan, G. & Peel, D. Finite Mixture Models. (John Wiley & Sons, 2000).

Post Author: admin


I'm Eric!

Would you like to get a custom essay? How about receiving a customized one?

Check it out