Declining Guillain-Barré Syndrome after Campylobacteriosis Control, New Zealand, 1988–2010

Food safety measures that lower incidence of campylobacteriosis might also prevent Guillain-Barré syndrome.

G uillain-Barré syndrome (GBS) is an autoimmune condition that affects the peripheral nervous system. Patients typically describe ascending weakness and sensory disturbance that evolve over several days; during this acute phase, approximately one third of patients require ventilatory support. The condition is generally selflimiting, but for 3%-10% of patients, it is fatal (1).
An estimated 40%-70% of patients with GBS had an infection before GBS onset; for 6%-39% of these patients, the infection affected the gastrointestinal system (2). Campylobacteriosis is the most commonly identifi ed antecedent infection; several studies have shown that in industrialized countries (Europe, North and South America, Japan, and Australia), Campylobacter spp. infection preceded GBS for 20%-50% of patients (3,4).
During 1980-2006 in New Zealand, incidence of campylobacteriosis steadily increased. The notifi cation rate in 2006 (379 cases/100,000 population) remains the highest national rate reported in the literature (5,6). In 2006, in response to this high incidence, New Zealand introduced an array of voluntary and regulatory interventions to reduce contamination of poultry with Campylobacter spp. (7). By 2008, the rate of campylobacteriosis notifi cations had dropped to 157 cases/100,000 population, a decrease of 59% over 2 years (7); this decline has persisted (8). Given the known association between Campylobacter spp. infection and GBS and the marked recent changes in reported rates of campylobacteriosis in New Zealand, we examined GBS hospitalization data for evidence of responsiveness to trends in campylobacteriosis incidence.

Identifi cation of GBS Incidence
Because GBS is a serious illness that nearly always results in hospitalization, hospitalization data provided the most accurate available measure of GBS incidence. We obtained national hospital discharge data for the 23-year period 1988-2010 in New Zealand. To estimate the casefatality proportion, we also obtained data on deaths from GBS for 1988-2008 (the most recent year available). Both datasets are collated and maintained by the New Zealand Ministry of Health.
Although hospitalization data are available for earlier years, we used 1988 as the starting point because that is when use of unique patient identifi ers, the National Health Index (NHI), became universal in New Zealand. Use of the NHI enables identifi cation and removal of repeat GBS hospitalizations for the same patient, thereby identifying the fi rst GBS hospitalization for each case (hereafter called GBS hospitalization), which provides an estimate of the number of incident cases of GBS.
We selected all cases from 1988 on that had International Classifi cation of Diseases, 9th and 10th Revisions, Clinical Modifi cation and Australian Modifi cation, codes for GBS (ICD-9 CM 357.0 and ICD-10 AM G61.0) recorded as the principal or additional diagnosis. Records of patients who had been transferred between hospitals were merged to create 1 hospitalization event. We identifi ed repeat hospitalizations for the current year and for previous years, i.e., case-patients with the same NHI number previously admitted in the same or a previous year. Some patients were readmitted before universal use of the NHI in 1988, so the calculation needed to take these estimated repeat hospitalizations into account. (See online Technical  Appendix Tables 1, 2, wwwnc.cdc.gov/EID/pdfs/11-1126-Techapp.pdf, for a description of how estimated repeat hospitalizations and incident cases were calculated.)

Identifi cation of Campylobacteriosis Incidence
Since 1980, campylobacteriosis has been a notifi able disease in New Zealand. Medical practitioners are required to report all identifi ed and suspected cases to the local medical offi cer of health. These data are in turn collated nationally by the Institute of Environmental Science and Research for the New Zealand Ministry of Health. We used published annual totals of notifi cations (9) as well as anonymized datasets of notifi ed cases. Most cases were culture confi rmed (>96% during 1997-2008 [7]), although the case defi nition also allows for cases epidemiologically linked to a confi rmed case.
Hospitalizations for campylobacteriosis are recorded in hospital discharge data, which are electronically available for a similar period. However, a specifi c diagnostic code for Campylobacter spp. infection was not introduced until July 1995. Hospitalizations for campylobacteriosis were defi ned as those with ICD-9 CM code 008.43 from July 1995 on and ICD-10 AM code A04.5 from July 1999 on. To create a dataset of incident cases, we included principal or additional diagnoses, merged records for those transferred with records from preceding hospitalizations, and removed repeat hospitalizations in the current and previous years.

Analysis of Hospitalizations for GBS after Campylobacteriosis
To assess the association between the 2 conditions, we investigated the incidence of GBS among patients hospitalized for campylobacteriosis. Because campylobacteriosis was only specifi cally identifi ed in hospitalization data from July 1995, this analysis focused on the period starting in July 1995. To allow a follow-up period for GBS cases to emerge, we continued the inclusion period through December 2008.
For those cases identifi ed, we fi rst analyzed the time from hospital admission for campylobacteriosis to admission for GBS. For epidemiologic purposes, the risk period for GBS after Campylobacter spp. infection is ≈2 months (10); neurologic signs of GBS usually develop 1-3 weeks after a preceding infection (3). In our dataset, a clear trend was seen toward a close temporal association between hospitalization dates: for most (34/35, 97.1%) patients, hospitalizations for GBS and campylobacteriosis were concurrent (patients were discharged with a diagnosis of both), or hospitalization for GBS occurred within 1 month of hospitalization for campylobacteriosis.
To assess the risk for GBS associated with campylobacteriosis, we calculated GBS hospitalization rates for comparison conditions, notably other infections that might be associated with an elevated risk for GBS. We used the GBS rate in the total New Zealand population as our reference rate for calculating age-standardized rate ratios for GBS after campylobacteriosis and other conditions of interest.
We also evaluated which age groups might be more vulnerable to development of GBS. To do so, we compared the age distributions of all patients hospitalized for GBS and those associated with campylobacteriosis with the age distributions for those with campylobacteriosis alone (hospitalized or with notifi ed case).

Statistical Analyses
Because of marked changes in campylobacteriosis disease incidence and some changes in case identifi cation during the 23-year study period, some outcomes were measured over a shorter time. The periods associated with implementation of the Campylobacter spp. control interventions used a baseline period similar to that used in a previous study (7). Data were analyzed by using Stata version 11.0 (StataCorp LP, College Station, TX, USA) and SAS version 9.1 (SAS Institute, Cary, NC, USA). CIs are given at the 95% level throughout. We used well-documented methods for calculating adjusted rates, rate ratios (RRs), and 95% CIs (11). Rates were calculated by using mean population estimates published by Statistics New Zealand (www.stats.govt.nz/browse_for_stats/population/ estimates_and_projections/national-pop-estimates.aspx) as denominators. To calculate age-standardized rates, we used the population age structure determined by the New Zealand 2006 Census of Population and Dwellings (www.stats.govt. nz/Census/2006CensusHomePage/classification-countstables/about-people/age.aspx).

GBS Incidence
This study identifi ed 2,056 fi rst hospitalizations for GBS that occurred during 1988-2010, resulting in an average rate of 2.32 hospitalizations/100,000 population/ year (online Technical Appendix Table 1). Incidence was not stable over the period of the study (Figure). The minimum recorded rate was 1.53 hospitalizations/100,000 population/year in 1989; the maximum was 2.93 in 2005. During 1989-2008, a total of 56 deaths from GBS were recorded; case-fatality proportion (56 deaths/1,873 cases) was 3.0%.

Changes in GBS and Campylobacteriosis Incidence
For 1988-2010, there was a signifi cant direct correlation between annual rates of hospitalization for GBS and annual rates of notifi cation of campylobacteriosis cases (Spearman ρ = 0.52, p = 0.012). During 1988-2006, incidence of campylobacteriosis notifi cations and of GBS hospitalizations increased ( Figure; online Technical Appendix Table 3). Subsequently, campylobacteriosis notifi cations then decreased markedly, and GBS hospitalizations decreased, although less dramatically. The fall in campylobacteriosis notifi cations followed the introduction of countrywide campylobacteriosis control measures focused on reducing contamination levels in fresh poultry meat (7).  Table 1 summarizes the changes between the 2 periods: 1) 2002-2006, the baseline period, when rising campylobacteriosis rates became an urgent public health concern, and 2) 2008-2010, the postintervention period, after implementation of wide-ranging control measures.
The transition year, 2007, was excluded.

GBS among Patients Hospitalized for Campylobacteriosis or Other Conditions
During 1995-2008, among the 8,448 patients hospitalized for campylobacteriosis, 35 were also hospitalized for GBS. The frequency distribution of time delays is shown in Table 2. These data show that most (29) of these 35 patients had diagnoses of GBS and campylobacteriosis at time of hospital discharge. Another 5 patients were hospitalized for GBS within 4 weeks of being hospitalized for campylobacteriosis. The time difference for the remaining patient was >1,500 days (this patient was excluded from subsequent analyses). This striking distribution further supports a causative association between campylobacteriosis and GBS in New Zealand.
We calculated the rate of GBS hospitalizations among the cohort of patients hospitalized for campylobacteriosis and compared this with rates of GBS hospitalization among other patient cohorts hospitalized for infectious diseases (Table 3). This analysis used the overall rate of GBS hospitalizations among the New Zealand population as a reference for calculating age-standardized RRs.
The age-standardized rate of GBS was 810.0 hospitalizations/100,000 person-years (95% CI 41.4-1,578.7) in the month after hospitalization for campylobacteriosis. The RR, compared with the rate of GBS hospitalizations among the New Zealand population, was 319.4 (95% CI 201. 5-506.4). This rate was markedly higher than rates for the other patient cohorts examined (Table 3).
Patients with GBS (median age 52.5 years) were signifi cantly older than those hospitalized for campylobacteriosis (median 41 years), who in turn were signifi cantly older than those with campylobacteriosis notifi cations (median 31 years) (Tables 4, 5). The age of the subpopulation of patients with GBS associated with campylobacteriosis was similar (median 54 years) to that of the total population with GBS.

Discussion
This study shows how the incidence of an acute infectious disease, campylobacteriosis, can infl uence incidence of a serious neurologic condition, GBS. At the population level, hospitalizations for GBS were signifi cantly correlated with notifi cations of campylobacteriosis for the same year. At the individual level, compared with rates for the New Zealand population as a whole, hospitalizations for campylobacteriosis were associated with an almost 320-fold increased risk for subsequent hospital admission for GBS in the next month.
Results also show that food safety measures to reduce contamination of fresh poultry meat with Campylobacter spp. not only reduced incidence of campylobacteriosis but also were associated with reduced incidence of GBS. In the 3 years after introduction of these control measures, campylobacteriosis notifi cations and hospitalizations decreased by ≈50%, and GBS hospitalizations dropped by 13%. These fi ndings suggest that in New Zealand, Campylobacter spp. infection may be responsible for ≈25% of GBS cases, which is consistent with data from other industrialized countries (3).
A recent systematic review (12) summarized attempts to quantify the association between campylobacteriosis and GBS incidence. There is general agreement that measuring GBS population rates is useful, for example, for monitoring vaccine adverse effects (13,14). However, to our knowledge, no similar population-based analysis of the relationship between GBS and campylobacteriosis has been conducted for other countries, probably because few countries collect similarly detailed national-level hospitalization data. An earlier population-based study in New Zealand did not show an association between notifi cations for campylobacteriosis and GBS incidence (15). However, that study was over a shorter period and did not use a correction factor to account for undetected repeat hospitalizations in the early years of the observation period, which would have made it harder to detect an association between incidence rates for the 2 conditions. Compared with global estimates, rates of GBS in New Zealand are high. In a review of reported GBS rates during 1980-2000, worldwide incidence varied between 1.0 and 1.8 cases/100,000 population/year (2). The average reported rate for New Zealand during this period was at the upper end of this range (1.8/100,000). A more recent study from the United States estimated that annual hospitalization rates for GBS varied between 1.65 and 1.79/100,000 during 2000-2004 (16). In New Zealand during the same period, the annual hospitalization rates varied between 1.8 and 2.7/100,000.
The 320-fold increased risk for GBS in the month after hospitalization for campylobacteriosis found in this study is higher than that previously reported. In a case-control study of GBS and potential antecedent infections in the United Kingdom, Tam et al. reported that persons with Campylobacter enteritis had a 38-fold increased risk that GBS would develop in the next 2 months (17). However, when they added a correction factor to account for underascertainment of campylobacteriosis, the risk increased to 60-fold. Similarly, a population-based study in Sweden estimated that patients with laboratory-confi rmed C. jejuni infection had a 100-fold increased risk that GBS would develop in the next 2 months (10). We used a 1-month risk period because the GBS cases we identifi ed subsequent to hospitalizations for campylobacteriosis were confi ned to this period. Using a 2-month risk period would have halved our estimated age-standardized RR, but the elevated risk would still be higher than that reported elsewhere.
The proportion of GBS cases attributable to preceding Campylobacter spp. infection estimated for New Zealand (≈25%) is within the range described elsewhere. Studies from other countries and regions have reported serologic evidence of previous C. jejuni infection in 13%-72% of GBS case-patients (18). A systematic review, based on 32 eligible studies, estimated that 31% of GBS cases were attributable to Campylobacter spp. infection (12).  Referent *GBS, Guillain-Barré syndrome; ICD, International Classification of Diseases; ICD-9, ICD 9th Revision; ICD-10, ICD 10th Revision; NA, not applicable. †Denominator population based on either 1) incident hospitalizations for specific condition (number of acute and arranged first overnight hospitalizations as principal or additional diagnosis); or 2) total New Zealand population person-years for July 1995-December 2008 for calculating the New Zealand population GBS rate. ‡First hospitalization of GBS either 1) among those with a previous hospitalization in the preceding 30 d and excluding those with concurrent diagnoses (numbers in parentheses); or 2) in the total New Zealand population for July 1995-December 2008.
§Rate per 100,000 person-years at risk. For GBS hospitalizations after specific conditions, monthly rate has been multiplied by 12 to convert to annual rate. ¶Standard population is population of New Zealand according to the New Zealand 2006 Census of Population and Dwellings (www.stats.govt.nz/Census/2006CensusHomePage/classification-counts-tables/about-people/age.aspx). #Excluding campylobacteriosis and salmonellosis.
The strength of the association with GBS may vary geographically, according to the neuropathic propensity of local Campylobacter strains. We would also expect the percentage contribution of preceding Campylobacter spp. infection to vary according to the incidence of this infection in the population and the incidence of other causal infections and exposures.
The results of our study suggest that risk for GBS may not be uniform for different degrees of campylobacteriosis severity. Our study found that risk for GBS was ≈1 in 1,690 (5 in 8,448) among patients hospitalized for campylobacteriosis and that ≈25% of GBS cases were caused by campylobacteriosis. On the basis of an annual incidence of ≈100 GBS cases, these data suggest that ≈42,000 cases of campylobacteriosis occur each year in New Zealand. Current estimates of total campylobacteriosis incidence are higher. Annual notifi cations remain at ≈7,000 cases. A study from the United Kingdom estimated that 9.3 cases of campylobacteriosis occurred in the community for every notifi ed case (19); a study from Australia estimated this number to be 10 (20). Applied to New Zealand, these multipliers suggest an incidence among the population of 65,000 to 70,000 cases per year. These fi ndings suggest that the causal association between campylobacteriosis and GBS is probably weaker for patients with less severe infections, who do not require hospitalization.
Analysis of the age distribution of patients with campylobacteriosis and GBS suggests that older age is a major risk factor for more severe outcomes (hospitalization and GBS) from this enteric infection. The rising incidence of GBS with increasing age in New Zealand is consistent with incidence observed in other countries (21).
One strength of this study is that it has been able to monitor a natural experiment in which campylobacteriosis incidence decreased by 50% within a few months, providing an unusual opportunity to assess the effect of this change on incidence of GBS. New Zealand's comprehensive recording of national hospitalization data and use of a unique patient number also provided us with a consistent base for estimating population rates of GBS over a prolonged period. Although the spectrum of GBS includes extremely mild cases, studies elsewhere indicate that only ≈3.0%-5.8% of patients with GBS are not hospitalized (22,23). In addition, patients with Campylobacter-associated GBS are believed to experience more severe disease (24,25), which would minimize the number of Campylobacter-associated GBS cases missed by this investigation.
One limitation of this study is the group used to compare risk for GBS: the total New Zealand population. A variety of conditions and events have been identifi ed as possible GBS triggers (1,24,(26)(27)(28)(29). Consequently, because it is not possible with current knowledge to identify a  reference patient population with no additional GBS risk factors, we considered that the total population provided the most appropriate reference rate. The association between campylobacteriosis and GBS in New Zealand needs further investigation. It will be useful to continue to follow the trends identifi ed here to assess the stability of the decrease in GBS, which will eventually give greater precision to the estimated contribution of campylobacteriosis. Ongoing monitoring of GBS should be included in the comprehensive surveillance of infectious diseases (30). The hypothesis that patients not hospitalized for campylobacteriosis have a lower risk for GBS should be tested by investigation of incidence of GBS among these patients.
Our fi ndings suggest the value of further research to identify other potentially preventable infectious causes of GBS. Table 3 shows a markedly elevated risk for GBS after hospitalization for infectious diseases in general. Investigating these associations in detail may identify other potentially preventable causes of GBS.
Findings of this study have relevant implications for food safety programs. Although GBS is rare, the toll it takes on the individual patient is often high (1). Even with treatment, 9%-17% of patients die or remain disabled (31), and repeat hospitalizations are common, representing ≈60% of total hospitalizations (online Technical Appendix Table  1). Almost half of all patients report ongoing diffi culties 3-6 years after GBS onset (32). Consequently, ongoing health care costs for each GBS patient are considerable. In New Zealand during 1988-2008, the GBS case-fatality proportion was 3.0%, and a recent article (33) estimated that 204 (13%) of 1,568 disability-adjusted life years for campylobacteriosis in New Zealand were caused by GBS.
This study shows that food safety programs that successfully lower rates of campylobacteriosis might have the additional benefi t of preventing GBS. This fi nding adds to the health and economic arguments for such control measures. The justifi cation for such interventions is particularly strong where a substantial proportion of human disease can be linked to a widely consumed food source, such as contaminated poultry products, as it is in New Zealand (7).