Rate of decline of antibody titers to pandemic influenza A (H1N1-2009) by hemagglutination inhibition and virus microneutralization assays in a cohort of seroconverting adults in Singapore

Abstract

Background: The rate of decline of antibody titers to influenza following infection can affect results of serological surveys, and may explain re-infection and recurrent epidemics by the same strain.

Methods: We followed up a cohort who seroconverted on hemagglutination inhibition (HI) antibody titers (≥ 4-fold increase) to pandemic influenza A(H1N1)pdm09 during a seroincidence study in 2009. Along with the pre-epidemic sample, and the sample from 2009 with the highest HI titer between August and October 2009 (A), two additional blood samples obtained in April 2010 and September 2010 (B and C) were assayed for antibodies to A(H1N1)pdm09 by both HI and virus microneutralization (MN) assays. We analyzed pair-wise mean-fold change in titers and the proportion with HI titers ≥ 40 and MN ≥ 160 (which correlated with a HI titer of 40 in our assays) at the 3 time-points following seroconversion.

Results: A total of 67 participants contributed 3 samples each. From the highest HI titer in 2009 to the last sample in 2010, 2 participants showed increase in titers (by HI and MN), while 63 (94%) and 49 (73%) had reduction in HI and MN titers, respectively. Titers by both assays decreased significantly; while 70.8% and 72.3% of subjects had titers of ≥ 40 and 160 by HI and MN in 2009, these percentages decreased to 13.9% and 36.9% by September 2010. In 6 participants aged 55 years and older, the decrease was significantly greater than in those aged below 55, so that none of the elderly had HI titers ≥ 40 nor MN titers ≥ 160 by the final sample. Due to this decline in titers, only 23 (35%) of the 65 participants who seroconverted on HI in sample A were found to seroconvert between the pre-epidemic sample and sample C, compared to 53 (90%) of the 59 who seroconverted on MN on Sample A.

Conclusions: We observed marked reduction in titers 1 year after seroconversion by HI, and to a lesser extent by MN. Our findings have implications for re-infections, recurrent epidemics, vaccination strategies, and for cohort studies measuring infection rates by seroconversion.

Figure 1

Scatterplot of titres from HI and MN assays. Random jitter was added to separate overlapping data points, and different coloured samples denote data from sample A (baseline) to sample C. The straight line in black is for the best fitting regression model, which has an R-squared of 0.51 (p<0.001).

Figure 2

Model goodness of fit of a multivariate ordered probit model to antibody titre distribution. In orange are HI assays (left column; Figures 2 A, 2 C and 2 E) and MN assays (right column; Figures 2 B, 2 D and 2 F) at three time points in 2009 (Figures 2 A and 2 B), April 2010 (Figures 2 C and 2 D) and September 2010 (Figures 2 E and 2 F), with 95% confidence interval error bars. In black is the posterior predictive distribution (mean and 95% credible interval). The ordered probit model accounts for age and sex as potential confounders, along with individual random effects and a temporal decay in antibodies, and uses the same θ thresholds at all time points. All the non-Gaussian distribution, and evolving shape of the distribution, are apparent, but the flexibility of the model formulation is able to account for both.

Figure 3

Modelled antibody trajectories by age category and gender on HI and MN assays. In orange is the posterior predictive distribution for individuals with age < 55 at three time points in 2009, April 2010 and September 2010, with 95% credible interval error bars (3A, 3B). In black is the posterior predictive distribution for individuals with age ≥ 55, with 95% credible interval error bars (3A, 3B). In blue is the posterior predictive distribution for males, with 95% credible interval error bars (3C, 3D). In pink is the posterior predictive distribution for women, with 95% credible interval error bars (3C, 3D). In red is the posterior predictive distribution for patients with any respiratory symptoms, with 95% credible interval error bars (3E, 3F). In grey is the posterior predictive distribution for patients without any respiratory symptoms, with 95% credible interval error bars (3E, 3F).

Figure 4

Estimated and observed proportions by age and sampling period for titres above respective cut-off points for (A) HI ≥40, and (B) MN ≥160. The orange bars with whiskers, which represent 95% confidence interval, indicate observed proportions stratifying by age and sampling periods. The black points with 95% credible interval error bars represent the estimated proportions adjusting by age and sampling periods.