Guillain-Barré syndrome following the 2009 pandemic monovalent and seasonal trivalent influenza vaccination campaigns in Spain from 2009 to 2011: outcomes from active surveillance by a neurologist network, and records from a country-wide hospital discharge database

The performance of our surveillance of GBS as a potential adverse reaction of influenza immunisations rests on the coincidence of time-related clustering of GBS cases occurring within the 42-day, high-risk post-immunisation period with a greater-than-expected increase in GBS incidence. Overall, monthly incidence of GBS, whether notified or registered country-wide, does not appear to reflect any impact of A(H1N1)pdm09 or seasonal vaccine campaigns. Our neurologist network failed to find a single GBS case following A(H1N1)pdm09 vaccination. The Hospital In-patient Registry suggested no increase in GBS monthly incidence during or post influenza immunisation campaigns.

Our study might have some limitations. The power of this study to detect an increase in incidence in vaccinated persons depends on the method of determination of the threshold, on the relative risk of GBS in vaccinated persons, and on the proportion of the population that received the vaccine. Since only a small part of the population was vaccinated with A(H1N1)pdm09, the relative risk associated with vaccination should have been very high to induce an overall incidence of GBS sufficiently high to exceed the threshold. However, we did not find any GBS case following A(H1N1)pdm09 vaccination in the 165,000 population vaccinated, despite an expected background incidence around 2.62 cases from December to March, which makes increased incidence associated with vaccination highly unlikely. Another limitation is the difficulty in adjusting for confounding from causes such as ILI-RTI or GTI, both of which are unlikely in our series in the light of the lack of such antecedents in six of our seven cases. Furthermore, potential underreporting might have happened, as has been suggested in connection with paediatric and medical departments in Sweden, where a similar GBS surveillance system was in place [27]. In our study, only one reported GBS case was not hospital-admitted but similar cases may have been overlooked. Underreporting in autumn 2009, linked to considerable media attention paid to influenza-related events during the influenza monovalent vaccine campaign, was probably lower than during other surveillance periods. Additionally, National Hospital In-patient Register has a proportion of false positive and false negative cases whose impact on this registry’s ultimate validity for estimating all GBS cases is uncertain [28]. Likewise, we assumed that background incidence estimated in 1999 would remain the same 10 years later, despite the potential change in infection transmissibility caused by the increase in population density in Spain over this period. Neither can exposure misclassification be ruled out since recall bias might have affected vaccination ascertainment. Our study also lacks power to study the individual effects of every vaccine adjuvant. Lastly, one of the post-2009-TIV Spanish cases had preceding ILI-RTI episodes, and two others had motor function potentially compromised by motor neurone disease. One could thus speculate that co-morbidity may have contributed to vaccination being indicated. High immunisation frequencies and improved information might have mimicked excess post-2009-TIV GBS among the surveillance population in both Spain and the United Kingdom [29].

Our results are consistent with a study by Burwen et al. [16] on GBS incidence after influenza vaccination among the Medicare population in 2009–2010, which failed to find an increase in GBS incidence above a critical limit established by GBS incidence in the previous five years. In contrast, our results differ from the modest risk excess reported in three pooled analyses [9–11]. However, the main design of these reviews, namely, self-controlled case series analysis and self-controlled interval-risk design, might imply a high risk of bias, i.e., recall bias for exposures predating GBS onset by more than 42 days and, in particular, incomplete adjustment for potential confounders like ILI-RTI or GTI and seasonal patterns of GBS onset [30, 31]. On the other hand, cohort studies either give lower relative risks [11, 15, 32] or fail to find any association [12–14].

Contrary to the findings of the ITANG study, which used a 176-patient case–control and case-series designs and reported modest but positive associations with 2010-TIV [33], we failed to observe any suspected impact of either the 2010 or the 2011 seasonal immunisation. Doubts nevertheless continue to surround ITANG’s capacity to remove residual confounding from the several-fold higher effect of the preceding infections, odds ratio (OR) = 23.8 and OR = 11.5 for ILI-RTI or GTI, respectively. Even so, the most important reason for suspecting a high risk of bias in the ITANG study is control selection towards low 2010-2011-TIV immunisation rates in hospital controls. ITANG controls were recruited from among patients with trauma, after eliminating those with chronic conditions, a selection strategy that would potentially exclude people in whom 2010-2011-TIV had been indicated.

One important aspect of our results regarding the four cases associated with the 2009-TIV and not with preceding infections is that they might match the intriguing association observed in the UK for the same exposure and time as reported by the VAESCO study. The VAESCO study reported an excess GBS risk for the unadjuvanted 2009-TIV in the UK [34], not seen in previous years using the same UK database [29]. The fact that such an effect was not observed in other seasonal influenza vaccines in 2010 or 2011 in Spain might point to a side-effect of the 2009-TIV due either to the A/Brisbane/59/2007(H1N1) or to the A/Brisbane/10/2007(H3N2) component of the 2009 TIV being removed from the 2010-2011-TIV.

Our study design, when compared to the other three procedures used to monitor or disclose potential causes of GBS, i.e., the spontaneous reporting of suspected adverse drug reactions, self-controlled case series and case–control analysis, raises public health service-related questions about preferential or alternative procedures when facing early detection of a potential excess risk of GBS. It would appear that methodological issues and logistic elements related to the question to be answered are paramount. Speed in activating surveillance resources may constitute a key issue. Self-controlled case-series analysis in New York based on a list of 150 active-reporting neurologists and 2495 passive-reporting neurologists based in neurology departments, was considered expensive [35]. Our system, relying on hospital neurology departments, took advantage of the National Health Care Services which, on a geographical residential basis, covered well-defined populations, was underpinned by a small number of neurologists and was conducted at no extra cost. Spain’s National Health Care system is operated by regional authorities and the AEMPS is a state-owned institution. The Spanish neurologist network was intensively active during the A(H1N1)pdm09 campaign, i.e., though the low monthly rates in 2010 and 2011 might have been related to underreporting. Reduction of surveillance to the November 2009-March 2010 period or express official support from the regional authorities for the purpose of incorporating public health missions in neurology departments might have improved GBS surveillance.

It would appear that GBS case–control studies using population controls and embodied in a hospital- and population-based network may be the best alternative for populations covered by publicly operated medical systems. This option does not exclude the addition of other procedures, such as self-controlled GBS case series, whether or not nested within an immunised cohort. At all events, knowledge of the GBS epidemiology in the study population is advantageous, particularly because asymmetry of exposure and confounder measurement in non-high-risk periods or among controls appears to be inevitable. Although A(H1N1)pdm09 monovalent vaccine is no longer in use, its strain has been incorporated in the seasonal vaccines of the following years, so that our study results are still of practical significance. All things considered, an effective influenza vaccine might reduce GBS incidence through reductions in ILI-RTI infections, a far more frequent antecedent to GBS than is vaccination, as shown by our results.