IntroductionGuillain Barré syndrome (GBS) is an autoimmune disorder of the peripheral nervous system triggered by autoantibodies formed in response to antigenic stimuli [1]. Antecedent exposures can include certain vaccinations (e.g., influenza) and viral or bacterial (especially Campylobacter) infections [2–6]. GBS is the most common cause of acute flaccid paralysis worldwide [1]; studies in Europe and North America report estimates of GBS incidence of 0.6 to 3.0 cases per 100000 person-years [1, 7]. GBS is associated with severe morbidity, with patients frequently requiring extended ICU stays and up to 67% experiencing at least one major complication [8, 9]. The economic cost is estimated to be $1.7 billion annually in the United States [10].
Campylobacter causes an estimated 1.3 million enteric illnesses annually in the United States, making it the most common bacterial cause of gastroenteritis [11]. C. jejuni accounts for most Campylobacter infections and has been estimated in various settings and countries to precede 20%– 31% of GBS cases with incidence estimated at 20-65 GBS cases per 100000 Campylobacter infections [2, 12–20]. However, recent estimates for US populations are not available [7, 20].
Determining the risk of post-Campylobacter GBS is challenging for several reasons. Campylobacter infection is often undetectable by the time GBS symptoms begin, because Campylobacter is typically shed for less than 2 weeks after onset of diarrhoea, whereas GBS symptoms typically present between 1 and 3 weeks after diarrhoea onset [2, 16, 20, 21]. In addition, due to mild symptoms or asymptomatic infection, many Campylobacter infections go undiagnosed, with an estimated 30 undiagnosed infections occurring for each laboratory-confirmed infection [11]. Diarrhoea can be mild [22], so infected persons may not seek care. Even if a stool sample is submitted, Campylobacter can be difficult to detect [23]. In the United States, surveillance for Campylobacter infection is conducted by the Centers for Disease Control and Prevention’s (CDC) Foodborne Diseases Active Surveillance Network (FoodNet), the foodborne disease component of the Emerging Infections Program (EIP) [24]; however, no routine surveillance for GBS exists [7].
An increased risk of GBS following vaccination with a specific formulation of the vaccine targeted at an H1N1 influenza virus was identified in 1976 [25, 26], though no significant increased risk was observed with subsequent seasonal influenza vaccines formulations [27–29]. However, when a novel influenza A (H1N1) virus similar to the type identified in 1976 emerged in 2009 [30–32], concerns about post-vaccination GBS arose, and CDC initiated a special EIP surveillance activity. This surveillance activity, conducted during the 2009–2010 novel influenza A vaccination campaign to assess the risk of post-vaccination GBS found no additional excess risk beyond typical that of seasonal influenza vaccines [33], and offered the opportunity for secondary analysis focused on post-Campylobacter GBS. This included extensive data collection on persons who were determined to not have GBS, providing a unique, well-characterized comparison group. Here, we report analysis of the association of GBS with laboratory-confirmed Campylobacter infection (the most specific measure for campylobacteriosis) and with diarrheal illness (the most sensitive, available measure) to estimate the fraction of GBS attributable to laboratory-confirmed Campylobacter infection or diarrheal illness. We also present estimated rates of post-Campylobacter GBS.
METHODSEIP GBS Surveillance ActivityWe used data from the EIP GBS surveillance activity conducted during the 2009-2010 novel influenza A vaccination campaign to analyse the association of GBS with diarrheal illness or laboratory-confirmed Campylobacter infection and to calculate the fraction of GBS attributable to diarrheal illness or laboratory-confirmed Campylobacter infection.
EIP GBS surveillance activity populationThe EIP includes ten sites and a population that is approximately representative of the U.S. population with respect to demographic and other health indicators, such as poverty (www.cdc.gov/ncezid/dpei/eip/). The catchment area for the GBS surveillance activity included 44.9 million persons. Data were collected between 1 October 2009 and 31 May 2010, yielding 22.9 million person-years under surveillance [33]. Possible GBS cases were identified by exhaustive, active, population-based case-finding to identify every resident of the catchment area presenting with symptoms possibly consistent with GBS. This case-finding was conducted through several avenues, including a network of clinicians (e.g., neurologists, clinical pharmacists, other providers), review of hospital admission and discharge data for the International Classification of Diseases-9-Clinical Modification code for GBS (357.0; acute infective polyneuritis), and monitoring of the Vaccine Adverse Events Reporting System (VAERS). For additional details, see Wise et al. [33].
GBS and non-GBS diagnosesData were collected by review of inpatient and outpatient medical records for all possible GBS cases identified with onset of symptoms during the surveillance period [33]. After data collection and review, all possible GBS cases were classified using the Brighton Collaboration criteria for GBS, a classification of diagnostic certainty [34]. Cases were classified as confirmed (meeting Brighton level 1 or 2 criteria) or probable (Brighton level 3 criteria) based on clinical, cerebrospinal fluid, and electrophysiologic criteria. We considered cases that did not meet the Brighton criteria for levels 1, 2, or 3 or cases in which an alternative diagnosis was reported as non-GBS controls.
Antecedent illnessInformation about signs, symptoms, and infections experienced in the 42 days before presentation, including diarrhea, influenza-like illness (ILI), upper respiratory tract infection (URI), and laboratory-confirmed Campylobacter infection, was collected for all of the reported possible GBS cases, including persons ultimately determined to not have GBS. GBS is known to be strongly associated with Campylobacter infection but less with other common causes of diarrheal illness, so we examined the association of antecedent illness with GBS diagnosis using five definitions that ranged from highly specific and less sensitive to highly sensitive and less specific for Campylobacter infection. The most specific, least sensitive definition was laboratory-confirmed Campylobacter infection. The most sensitive, least specific was any diarrheal illness, which, as described above, was used because Campylobacter infection is usually not laboratory-confirmed. Three additional definitions of intermediate specificity and sensitivity included diarrhoea without ILI, diarrhoea without URI, and diarrhoea without either ILI or URI. These were used because ILI and URI can also precede GBS and can sometimes include diarrhoea [12].
FoodNetFoodNet, the foodborne diseases component of the EIP, is a collaboration among CDC, ten state health departments, the U.S. Department of Agriculture’s Food Safety and Inspection Service (USDA-FSIS), and the Food and Drug Administration (FDA). It conducts active, laboratory-based surveillance for selected pathogens transmitted commonly by food, including Campylobacter, and publishes annual estimates of incidence. The FoodNet population is similar though not completely identical to the 2009-2010 GBS surveillance population. Based on 2010 U.S. Census data, about 18% of the FoodNet surveillance population resided in areas not included in the EIP catchment and about 15% of the EIP GBS surveillance population was not included in the FoodNet catchment. We used FoodNet data on laboratory-confirmed Campylobacter infections reported from 15 September 2009 to 14 September 2010. Since the EIP GBS surveillance activity did not cover a full year, we used FoodNet data from 2009-2010 on the timing of laboratory-confirmed Campylobacter infection in our extrapolation from 8-month to 12-month estimates. Thus, we calculated the proportion of laboratory-confirmed Campylobacter infections reported to FoodNet that occurred during 15 September 2009 – 15 May 2010, a period shifted 2 weeks earlier than the GBS surveillance activity, to account for an average 2-week lag between onset of Campylobacter-related diarrhoea and onset of GBS.
Statistical analysesConfirmed and probable GBS cases (Brighton 1–3) were compared to non-cases to determine whether antecedent illness, as determined using the five definitions detailed above, was more common in cases. We calculated odds ratios (OR) to evaluate the association between each definition of antecedent illness and GBS, and we used these OR to estimate the attributable risk (AR) [35].
The AR estimates in turn were used to estimate the number of post-Campylobacter and post-diarrheal GBS cases that occurred in the EIP GBS surveillance activity population during the surveillance period. Next, incorporating the national estimate of Campylobacter incidence data, we estimated national rates of post-Campylobacter (post-diarrheal) GBS in the United States using each of the five definitions of antecedent illness. All analyses were performed in SAS 9.3 (Cary, NC), Microsoft Excel, or the R Package, epiR.
For sensitivity analysis, we also used a more specific definition of GBS limited to confirmed GBS (Brighton levels 1 and 2). We also repeated analyses excluding the 11% of patients referred for possible GBS who had a previous history of GBS.
DISCUSSIONHigh quality, comprehensive, population-based, active surveillance data are rarely available for GBS, which, though uncommon, is responsible for high morbidity and economic burden [8–10]. We conducted a secondary analysis of GBS surveillance data collected during the 2009-2010 novel influenza A vaccination campaign to generate contemporaneous estimates of the burden of GBS attributable to Campylobacter infection in the United States; the primary analysis demonstrated that the risk of GBS following novel H1N1 vaccination was extremely low, and not greater than what is typically observed for seasonal influenza vaccines. We estimate that 8.2-12.3% of GBS is attributable to antecedent Campylobacter infection, with 433 to 650 cases of GBS occurring annually in the United States (32.8-49.2 per 100000 Campylobacter infections) that are attributable to antecedent Campylobacter infection. Although attributable risk estimates are at the lower end of the range of previous estimates for the US and other developed countries, the incidence estimates are in the mid- to upper- range [17–20].
The EIP GBS surveillance activity provided a unique opportunity to investigate the association between Campylobacter and GBS. Strengths of the project include detailed health history collected through intensive, active, population-based surveillance not only from individuals who met the GBS case definitions but also from a comparison group. Given that Campylobacter infection is usually not laboratory-confirmed (only six laboratory-confirmed cases were reported in the project), and diarrhoea often resolves before GBS symptom onset [2, 16, 20, 21], the collection of signs and symptoms in the 42 days prior allowed exploration of multiple definitions of varying sensitivity and specificity to represent antecedent Campylobacter illness. Of note, although the catchment areas of the EIP GBS surveillance activity and FoodNet did not perfectly overlap and Campylobacter incidence estimates were geographically contingent, the low proportion of mismatch in the catchment areas would not be expected to lead to a large difference in our results.
Campylobacter is the most common bacterial cause of domestically-acquired acute gastroenteritis in the US [11, 36]. With rare exceptions [3, 37], the other top three acute gastroenteritis pathogens (norovirus, Salmonella, and Clostridium perfringens) have not been consistently associated with GBS. However, the less specific but more sensitive definitions of antecedent Campylobacter illness based on diarrheal symptoms may have misclassified other diarrheal infections that are rare antecedents of GBS. The impact of these biases is hard to predict. On one hand, using diarrhoea as a proxy for campylobacteriosis should overestimate antecedent illness in both cases and controls, leading to underestimation of the association between Campylobacter infection and GBS. On the other hand, to the extent that other diarrheal syndromes are truly associated with GBS, attributing them to Campylobacter would lead to an overestimate of the post-Campylobacter association.
A limitation of this project is that some patients with Campylobacter infection may not have reported diarrhea. For example, one patient with culture-confirmed Campylobacter infection did not report diarrhoea and therefore was not captured by the definition of antecedent diarrheal illness. However, GBS diagnosis (case versus non-case) would not have influenced testing for Campylobacter or report of diarrhoea in the previous 42 days because these occurred before the onset of the symptoms that led to reporting of possible GBS. In addition, identification of Campylobacter infection was limited to reported symptoms and clinical culture; serological testing was not performed. This may have led to underreporting of Campylobacter infection, thus underestimating the reported association.
Campylobacteriosis was not nationally notifiable at the time of the EIP GBS surveillance project. Therefore, a major strength of using FoodNet surveillance data for the annual incidence of Campylobacter infection in the United States is that the data were collected through active laboratory-based surveillance, which estimates infections and incidence rates more accurately than passive surveillance. The estimated annual incidence of campylobacteriosis was generated using 2006 data, while the EIP GBS surveillance project covered an 8-month period during 2009 to 2010. This is unlikely to have substantially affected our results, as the incidence of Campylobacter infection remained relatively stable between 2006 and 2010 [38].
This analysis provides updated estimates related to GBS cases following Campylobacter infection in the United States. Post-Campylobacter GBS tends to be more severe than GBS following other antecedent events, with worse outcomes and slower recovery [14]. Campylobacter infections in the US have an estimated economic burden of ($1.9 billion), over half which is attributed to GBS-related morbidity and mortality [39]. Efforts to decrease Campylobacter infections, a priority of the Food Safety Modernization Act, would likely contribute to a decrease in GBS, specifically the most severe GBS cases, thereby substantially mitigating morbidity and mortality associated with Campylobacter infection.