Submitted by: Bruce L Innis, MD and Philippe Boutet, PhD, GlaxoSmithKline Vaccines, King of Prussia, PA, USA and Wavre, Belgium

Melen et al. [1] sought to evaluate whether coinciding pandemic influenza A/H1N1pdm09 infection contributed together with the vaccination campaign using AS03-adjuvanted Pandemrix™ (A/H1N1pdm09 vaccine, GlaxoSmithKline, Belgium) to the increased incidence of childhood narcolepsy in Finland observed beginning in 2010. Based on an analysis of anti-influenza A (A/H1N1pdm09) NS1 antibodies detected by western blot, they concluded there was no serological evidence of influenza A/H1N1pdm09 virus infection as a contributing factor. However, their data support an alternative interpretation that at least 71% of the narcoleptic patients (32/45 in Table S2) had serologic evidence consistent with, though not diagnostic of, past infection with A/H1N1pdm09. These patients had an anti-NS1 WB reciprocal titer ≥ 100, the assay minimum threshold detection level evident for the sera of confirmed A/H1N1pdm09 cases (Figure 2). The sera were collected in 2011-2012, after the patients’ narcolepsy had been diagnosed. Consequently, we find the evidence presented to be inconclusive as to whether narcoleptic patients had sustained A/H1N1pdm09 infection before or after they were vaccinated and the title of their report to be misleading.

As discussed below, there are two errors of logic in the analysis. First, a serologic test can reliably exclude past infection only if the antibody response to be measured is known to persist above a detection titer for a sufficient time period to permit retrospective analysis and if the test is highly specific. The western blot assay for retrospective identification of A/H1N1pdm09 infection met neither of these performance criteria. Second, the authors reasoned that finding geometric mean titers (GMT) of anti-NS1 (A/H1N1pdm09) antibody that were 2.7-fold lower in narcoleptic patients than in age-matched children bled 6-7 years earlier (418.2/156.1 as shown in Table 2) was evidence that narcoleptic patients had no exposure to A/H1N1pdm09 infection. This is an unwarranted inference, as the western blot test is not specific for influenza A subtypes and the cumulative influenza A virus exposure history of the narcoleptic patients sampled in 2011/2012 cannot be assumed to be equivalent to that of children sampled in 2004/2005.

The authors found that the sensitivity of the western blot assay to detect acute A/H1N1pdm09 infection was 89% (25/28 patients who had A/H1N1pdm 09 infection confirmed by virus detection in respiratory secretions, as shown in Table S3) using a criterion of a 2-fold rise in titer to a level ≥1:600 or a markedly elevated titer with no rise in 14-21 day paired serum specimens. However, as the authors acknowledged, the anti-NS1 assay was non-specific, because 23 of 25 positive serum pairs evidenced similar fold rises in anti-NS1 titer using A/Udorn/72 (H3N2) antigen. Their data support that a positive test for anti-NS1 (A/H1N1pdm09) is consistent with but not diagnostic of A/H1N1pdm09 infection.

Their analysis of serology in 45 narcoleptic patients used serum collected in 2011. As the first pandemic wave of A/H1N1pdm09 infection in Finland peaked in September 2009, these specimens were obtained at least one year or more after any H1N1 infection that may have occurred coincident with the national immunization campaign in which Pandemrix™ was administered. Unfortunately, the authors did not obtain one year convalescent serum specimens from control patients with confirmed A/H1N1pdm09 infection to establish the sensitivity of their western blot test to detect anti-NS1 at this time point. As it is likely that anti-NS1 levels diminished over time, this is an additional limitation of the serology they performed.

The authors found a 94% prevalence of anti-NS1 (A/H1N1pdm09) antibodies with a GMT of 1:418 in serum specimens that were collected in 2004/2005 (before the A/H1N1pdm09 emerged in humans) from 50 children age-matched to the 45 narcoleptic patients (47/50 specimens with reciprocal titer ≥100 in Table S4). In contrast, the prevalence and GMT of anti-NS1 (A/H1N1pdm09) antibodies in late convalescent serum specimens from narcoleptic patients was 71% and 1:156. Although this panel of 2004/2005 serum specimens is presented as a control, given the lack of test specificity for antibodies to A/H1N1pdm09 NS1, the clinical significance of the observed contrast in prevalence and GMT is uncertain.

Given the serologic test’s limitations with respect to observed sensitivity and specificity, the authors seem to have set a criterion of equal anti-NS1 titers for A/H1N1pdm09 and H3N2 (A/Udorn/72) as defining past A/H1N1pdm09 infection. Only 2 narcoleptic patient serum specimens (4%) met this criterion. The authors interpreted this observation as providing “indirect evidence that A/H1N1pdm09 virus infection was not common among our narcoleptic patients.” However, the criterion they set was never validated and the authors present no evidence supporting the use of their criterion. In fact, the relative difference in anti-NS1 titers between subtypes in a narcoleptic patient’s late convalescent serum specimens has unknown significance. Even if the test was sufficiently standardized to conclude on differences in anti-NS1 concentration between H1N1 and H3N2 subtypes, there would be no surprise in finding that anti-NS1 (H3N2) levels were higher. It is reasonable to suppose that children have been more frequently infected with H3N2 than with H1N1, given the higher antigenic drift rates of the H3N2 subtype. Higher subtype infection rates would lead to higher concentrations of circulating antibody.

Without explanation, the authors have dismissed the 71% prevalence of anti-NS1 (A/H1N1pdm09) antibodies in late convalescent serum specimens of narcoleptic patients as lacking significance. However, this observation cannot be set aside. Based on the data they have generated, a positive test for anti-NS1 (A/H1N1pdm09) is consistent with, but not diagnostic of, past infection. Given the data and the limitations of the methodology used, a more accurate title for the manuscript would have been “The serological evidence does not exclude influenza A/H1N1pdm09 virus infection as a contributing factor in childhood narcolepsy after Pandemrix™ vaccination campaign in Finland” as neither the title of the manuscript nor its conclusion are supported by the data from the study.

This distinction matters because a recent publication [2] presents evidence that CD4 T-cells from narcoleptic patients recognize a peptide encoded in hypocretin (the hormone deficient in narcolepsy) that is mimicked by a similar peptide uniquely encoded in the A/H1N1pdm09 hemagglutinin protein. This work supports the hypothesis that certain kinds of exposures to A/H1N1pdm09, whether via vaccination, infection, infection then vaccination, or vice-versa, may in rare individuals with unknown predisposing factors trigger immune activation of auto-reactive T-cells that could promote attack on hypocretin-producing neurons in the hypothalamus. The sequence of events necessary to render a person narcoleptic after exposure to the A/H1N1pdm09 hemagglutinin protein is unknown and warrants further evaluation.

Trademark statement: Pandemrix is a trademark of GlaxoSmithKline group of companies

References
1. Melen K, Partinen M, Tynell J, Sillanpaa M, Himanen S-L, et al. (2013) No Serological Evidence of Influenza A H1N1pdm09 Virus Infection as a Contributing Factor in Childhood Narcolepsy after Pandemrix Vaccination Campaign in Finland. PLoS ONE 8(8): e68402. doi:10.1371/journal.pone.0068402
2. De la Herrán-Arita AK, Kornum BR, Mahlios J, Jiang W, Lin L, Hou T, Macaubas C, Einen, M, Plazzi G, Crowe C, Newell EW, David MM, Mellins ED, Mignot E, 2013. CD4+ T Cell Autoimmunity to Hypocretin/Orexin and Cross-Reactivity to a 2009 H1N1 Influenza A Epitope in Narcolesy. Sci Transl Med 5:216 doi: 10.1126/scitranslmed.3007762