Guillain-Barré Syndrome Incidence in a Large United States Cohort (2000–2009)Shui I.M.a, b · Rett M.D.a · Weintraub E.e · Marcy M.f · Amato A.A.c · Sheikh S.I.c · Ho D.c · Lee G.M.a, d · Yih W.K.a · for the Vaccine Safety Datalink Research Team
aDepartment of Population Medicine, Harvard Medical School and Harvard Pilgrim Health Care Institute, bDepartment of Epidemiology, Harvard School of Public Health, cDepartment of Neurology, Brigham and Women’s Hospital and Harvard Medical School, and dDepartment of Laboratory Medicine and Division of Infectious Diseases, Children’s Hospital, Boston, Mass., eImmunization Safety Office, Division of Healthcare Quality and Promotion, Centers for Disease Control and Prevention, Atlanta, Ga., and fKaiser Permanente Southern California, Pasadena, Calif., USA
Background/Aim: We describe the incidence of Guillain-Barré syndrome (GBS) in a large United States cohort. Methods: Between 2000 and 2009, we identified visits with an ICD-9 code for GBS (357.0) from all persons with continuous enrollment for at least 1 year. The primary case definition was restricted to emergency department and inpatient visits. We calculated age-standardized rates and used multivariate Poisson regression to assess variation in rates by sex, age, season and year of diagnosis. We tabulated descriptive characteristics and the positive predictive value (PPV) for a subset of the visits with available medical record review. Results: 1,619 visits with the GBS ICD-9 code were identified from 50,290,898 person-years of observation. After considering the PPV (55%) for record-reviewed visits, the age-standardized incidence rate was approximately 1.72/100,000 person-years. The rate was 40% higher for males and increased by 50% for every 10-year increase in age. The rate was 15% higher in winter and spring compared with summer. Rates were higher in more recent years. Conclusions: GBS rates are higher in males and increase considerably with age. The potential reasons for differences in rates by season and the increased rates in more recent years should be further investigated.
Copyright © 2012 S. Karger AG, Basel
Guillain-Barré syndrome (GBS) is a rare and serious autoimmune disorder of peripheral nerves . A comprehensive review estimated GBS incidence rates in North America and Europe to be between 1–2 per 100,000 person-years . The rate increases with age and male sex [2,3,4]. Strong seasonal associations or clear secular trends in GBS rates have not been documented [2,5]. An infectious illness within the preceding 6 weeks, most frequently gastroenteritis or respiratory tract infection, is often reported in patients diagnosed with GBS [2,5]. Although current scientific evidence does not support a causal association for most vaccines and GBS, concern remains due to the documented increase of approximately 1 case per 100,000 vaccinations after the swine influenza (H1N1) vaccine in 1976 . Several studies have assessed associations between other vaccines and GBS and the majority did not find evidence for an increased risk [7,8,9,10,11,12,13]; three studies found a very small increase equivalent to approximately one additional case per million vaccinations with seasonal or 2009 H1N1 monovalent influenza vaccines [14,15,16]. Given the severity of GBS and continued interest in whether there is an increased risk of GBS following vaccination , we conducted a comprehensive study in the Vaccine Safety Datalink (VSD) population to better characterize the epidemiology and baseline incidence rates of GBS.
Materials and Methods
The VSD is a collaboration between the Centers for Disease Control and Prevention and eight managed care organizations (MCOs) with the overall goal of evaluating vaccine safety. The design and data collection methods of the VSD have been described in detail . Annually, over 9 million persons are enrolled and computerized data on demographics, immunizations, and medical encounters, including ICD-9 diagnosis codes, are collected. Each participant has a unique study ID, which prevents duplicate inclusions if the participant is seen in two settings for the same episode [e.g. emergency department (ED), then admitted to inpatient]. Access to participants’ medical records is also available. Seven of eight MCOs within the VSD were included in this study: Group Health Cooperative (Seattle, Wash., USA), Harvard Pilgrim Health Care and Harvard Vanguard Medical Associates (Boston, Mass., USA), HealthPartners Research Foundation (Minneapolis-St. Paul, Minn., USA), Kaiser Permanente of Colorado (Denver, Colo., USA), Kaiser Permanente of Northern California (Oakland, Calif., USA), Northwest Kaiser Permanente (Portland, Oreg., USA), and Kaiser Permanente of Southern California (Pasadena, Calif., USA). Children aged less than 1 year were followed from the earliest date of enrollment; otherwise, all persons in the cohort with at least 1 year of continuous enrollment during 2000–2009 were included.
We identified visits for GBS from the automated data records using the ICD-9 code 357.0 (acute infective polyneuritis) in any setting of diagnosis (inpatient, ED or outpatient). If a member was seen in more than one setting within a 2-day period, we applied the following algorithm. If a GBS code was given during an ED or outpatient visit and a subsequent code was given for an inpatient admission, we classified it as inpatient. If the first code was given in the ED or inpatient setting and a subsequent code was given in the outpatient setting, we classified the visit as an ED visit or inpatient visit, respectively. To prevent capturing follow-up visits for management of existing GBS, we restricted the study to the first ICD-9 code given in a 1-year period. The primary case definition was further restricted to ED and inpatient visits. A secondary case definition also included outpatient visits.
Medical record review was not available for all visits with the GBS ICD-9. We were able to access record reviews for a subset of the visits, which occurred within 42 days of a seasonal trivalent inactivated influenza vaccine, that were previously conducted for a separate study assessing the risk of GBS following the 2009 H1N1 influenza vaccine. Information was abstracted using an instrument based on the current Brighton Collaboration definition  and visits with potentially acute onset GBS underwent final review and adjudication by a neurologist with expertise in GBS. Separate criteria were used for GBS and Fisher syndrome, a GBS variant (table 1), and cases were categorized into four levels of certainty (table 2).
|Table 1. Case definition criteria|
|Table 2. Levels of diagnostic certainty for GBS or FS – required data|
Using the automated data, we calculated the rates for visits with a GBS ICD-9 code and 95% confidence intervals (CI), stratified by sex and the following age groups (years): 0–4, 5–17, 18–24, 25–49, 50–64, and 65 or older. The rates were standardized to the US Census 2000 population. Descriptive information on the clinical characteristics and the positive predictive value (PPV) of the ICD-9 code 357.0 were tabulated for the subset of record-reviewed visits. We also applied the PPVs from the record-reviewed samples to the incidence rates based on the automated data to provide an estimate for the rate of confirmed or probable cases. Multivariate Poisson regression assessed variation in rates with respect to sex, age group, year of visit, or season (winter: December to February, spring: March to May, summer: June to August, fall: September to November), while adjusting for site. All participating institutions granted Institutional Review Board approval.
One hundred and nine visits with a GBS ICD-9 code in the 42 days following seasonal influenza vaccine were identified for medical record review and records for 99 of these visits were obtained. There were 38 confirmed/probable cases: 28 confirmed (Brighton levels 1–3) and 10 probable (without sufficient clinical or laboratory evidence to meet the Brighton levels 1–3) (fig. 1). Most of the confirmed or probable cases (29/38, 76%) were identified from the inpatient or ED settings. The PPV for the GBS visits identified in these settings (n = 53) was more likely to be confirmed or probable (PPV = 55%) compared with the outpatient setting (n = 46; PPV = 20%). The predictive values did not vary significantly by age or sex. When we restricted the study to confirmed cases the PPV decreased to 28% (all settings) and 47% (ED and inpatient).
|Fig. 1. Classification of 38 confirmed GBS and Fisher syndrome (FS) cases.|
Antecedent infection was present in 22/38 (58%) of the confirmed/probable cases; infections were most often upper respiratory/influenza-like or gastrointestinal illness. Most cases had documented treatment; 74% were treated with intravenous immune globulin and 21% received plasma exchange; 13% required intubation and mechanical ventilation. During a median follow-up time of 6.5 months, there were 2 deaths in cases aged 59 and 81 years, which occurred 249 and 160 days after onset. Ten (29%) cases had reached a full/nearly full recovery, while 5 (15%) cases still had chronic/residual symptoms. The remaining 21 (62%) patients had been discharged to home/rehab or had reached a stable clinical plateau and their final outcome was unknown at the time of record review.
For the primary definition, 1,619 visits with the GBS ICD-9 code were identified in the ED and inpatient settings from a total of 50,290,898 person-years of observation. Table 3 shows incidence rates of the visits for GBS (per 100,000 person-years) and 95% CI by age group and sex. Age-standardized rates were 3.13/100,000 person-years for the total population (95% CI: 2.98–3.28), 2.64/100,000 person-years (95% CI: 2.45–2.84) for females and 3.70/100,000 person-years (95% CI: 2.46–3.95) for males. The incidence of GBS increased with age and male sex. It is important to note that these rates were computed using visits identified by ICD-9 codes only and were not based on record-reviewed data. However, after applying the PPV (including probable cases) of 55% for the ICD-9 codes from our sample of record-reviewed cases in the ED and inpatient setting, the corresponding age-standardized rates (per 100,000 person-years) were approximately: 1.72 (total population), 1.45 (females) and 2.04 (males). Further restricting to the PPV for confirmed cases (47%) yielded approximate rates of 1.47 (total population), 1.24 (females), and 1.74 (males). The incidence rates including the outpatient clinic setting were substantially higher, but the PPV for the outpatient setting was considerably lower.
|Table 3. Incidence rate (IR) of visits with an ICD-9 code for GBS per 100,000 person-years by age and sex from automated data (not record reviewed)|
There was a 50% increased rate for each 10-year increase in age and a 40% increased rate for males compared with females (table 4). Rates for visits because of GBS varied significantly by season and were approximately 15% higher in winter and spring when compared with summer and fall (table 4). Although a strong secular trend was not observed, rates for visits because of GBS did fluctuate by year of diagnosis and were lowest during the years 2003–2004 and highest in the most recent years (2008–2009) of the study period (table 4). A sensitivity analysis including the outpatient clinic visits yielded qualitatively similar results (data not shown).
|Table 4. Multivariate Poisson regression for predictors of visits with an ICD-9 code for GBS|
Our study provides a comprehensive and current assessment of the GBS incidence rates and epidemiology in a large cohort of MCO patients in the USA. Knowledge of accurate background rates for rare diseases such as GBS is useful in characterizing the epidemiology of the disease; such rates have been frequently used for providing a baseline comparison group for the assessment of vaccine safety during mass immunization programs [7,8,17,19,20]. In addition, postlicensure surveillance for adverse events after vaccines and drugs commonly relies on computerized ICD-9 codes to identify potential outcomes; our study provides rates from both the computerized data alone and considers the PPV of a sample of the cases. The age-standardized rate from computerized data was 3.1/100,000 person-years for the total population; the rate increased by 50% for every 10 years in age and was 40% higher in males than females. The increase in GBS rates with advancing age is well documented as are the higher rates of GBS among males, although the biologic reason behind the increased rates of disease in males is unclear [2,5,10,21]. After applying the PPV of 55% (including probable cases), the age-standardized incidence rate of GBS was approximately 1.7/100,000 person-years using our primary case definition. This is comparable to other US and European GBS rates reported in the literature [2,5].
In our study, the PPV varied by setting of diagnosis and by whether we excluded probable cases. The PPV for the primary case definition, including probable cases, was 55%; however, it was only 28% when we included outpatient clinic visits and restricted the study to only Brighton-confirmed cases. Several other studies have reviewed the PPV for GBS identified from automated data sources using the ICD-9 code 357.0 and found a range of values. The PPV was 30% in a study of hospital discharge databases in Tennessee during 2002–2003 , 71% in a study using Vermont hospital discharge databases from 1980 to 1985 , and 78% in a study conducted in a large MCO in California that identified cases from outpatient, ED, and inpatient visits with the ICD-9 code as well as text write-in diagnoses . The variability in PPV could be attributable to differences in the study populations (including coding practices by the treating physicians), inclusion criteria for case ascertainment, and differences in the criteria used for case confirmation. Some studies identified cases from hospital discharge records only [4,22], while our study and the one by Klein et al.  were cohort studies based on MCOs. Both of the latter studies had specific enrollment criteria required for inclusion and restricted diagnoses to the ‘first given in a specific period’ to better identify incident GBS and to eliminate duplicate visits for the same episode (e.g. hospital transfers). Finally, criteria for case confirmation varied; our study and two others [4,22] used standardized criteria (e.g. Brighton levels 1–3 or the National Institute of Neurological and Communicative Disorders and Stroke criteria) , but one study  used less stringent criteria (neurologist statement). Few studies, including ours, have been able to assess the sensitivity of the ICD-9 code 357.0, but because GBS is a severe disease that requires medical attention, we expect that most true acute cases should be captured in our database. One northern Italian study compared cases identified by the hospital discharge codes and a well-established GBS registry, which was also supplemented with data from a regional case-control study, and found a sensitivity of 91% .
We observed approximately 60% of the record-confirmed GBS cases had evidence of antecedent infection, which is consistent with several studies [2,5,10,26]. Most (70%) of these cases had evidence of influenza-like illness or upper respiratory infection, while 13% had gastrointestinal infections; however, this pattern must be interpreted with the caveat that all cases selected for record review occurred during the influenza season. Campylobacter jejuni, a gastrointestinal infection, is the most commonly identified antecedent infection among GBS cases ; however, recent studies have also found an increased risk of GBS in the period immediately following influenza-like illness [10,26,27]. Our results show a higher proportion of patients treated with intravenous immune globulin (74%) compared with older studies (30–40%) , reflecting the shift in treatment preference following randomized control results which show that IgIV has similar efficacy to plasma exchange and is more convenient to administer . The clinical course of GBS in our study was in agreement with that from other case series .
We observed significant seasonal variation, with an approximately 15% higher rate of visits for GBS reported during the winter and spring when compared with the summer and fall. The reason for this seasonality is unclear, but may be consistent with a higher incidence of upper respiratory infections or influenza-like illness during these months. A study by Stowe et al.  conducted in a United Kingdom cohort also found that the seasonal distribution of reported GBS was higher in the winter months, and other studies have indicated some evidence for more cases of GBS in the colder months [3,28,29]. However, other studies did not show clear seasonal patterns [2,21]. Unlike the two studies which examined GBS rates in the last decade and did not find evidence for a secular trend [10,30], we found that GBS rates increased over the study period. The reasons for this increase are unclear, but could be due to changes in coding practices, diagnosis, or an actual increase in GBS rates over the years, and deserve further investigation. We were unable to address regional differences in GBS rates by MCO site due to sparse data.
A main limitation for our study is that the analysis was primarily based on visits with ICD-9 codes for GBS from automated data. Medical record review was available for a subset of the visits which had occurred within 42 days after influenza vaccination. Because this was a convenience sample and not a random sample of the total population, there is a possibility that the PPV may not be representative of all visits. For example, if there were a heightened awareness of GBS following influenza vaccination, physicians might be more likely to code a rule-out diagnosis which would tend to make the PPV in our subset lower than that of the general population. We did not observe a difference in predictive value by age or sex. When we applied the PPV from the subset of cases, which did undergo review, the adjusted rates were comparable with those in recent literature reviews, and the patterns we observed for age- and sex-stratified rates were also in accordance with the current available knowledge.
Major strengths of this study were the presence of reliable population denominator data to calculate rates, and relatively unbiased case ascertainment. In comparison hospital-based review studies often have an unclear base population from which to estimate denominators and may also suffer from ascertainment and referral bias. In addition, we described the clinical characteristics and the PPV of GBS diagnostic coding for a subset of the visits identified. Our medical record process was based on established Brighton collaboration criteria and adjudicated by a neurologist with expertise in GBS.
GBS incidence rates from our study are similar to those observed in other recent US and European studies, are higher in males, and increase considerably with age. The majority of record-confirmed cases were preceded by an infection. The secular trend in the rates over the last decade and the potential for seasonality should be further investigated.
Vaccine Safety Datalink Research Team Coinvestigators: Kaiser Permanente Vaccine Study Center, Northern California Kaiser Permanente Division of Research, Oakland, Calif., USA: Roger Baxter, MD; Center for Health Research, Kaiser Permanente Northwest, Portland, Oreg., USA: Allison L. Naleway, PhD; Group Health Center for Health Studies, Group Health Cooperative, Seattle, Wash., USA: Lisa Jackson, MD; Institute for Health Research, Kaiser Permanente Colorado, Denver, Colo., USA: Simon Hambidge, MD, and HealthPartners Research Foundation, Minneapolis, Minn., USA: James Nordin, MD.
The authors would also like to acknowledge: Stephanie Irving (MFC), Paula Ray (NCK), Rong Li (HAR), Sharon Greene (HAR), Tracy Lieu (HAR), Richard Platt (HAR), Patricia Kennedy (HAR), and Lingling Li (HAR).
This study was funded through a subcontract with America’s Health Insurance Plans (AHIP) under contract 200-2002-00732 from the Centers for Disease Control and Prevention (CDC).
The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention nor that of America’s Health Insurance Plans.
The authors declared no conflicts of interest and have no disclosures to make.
Irene M. Shui, MPH, ScD
677 Huntington Avenue
Boston, MA 02115 (USA)
Tel. +1 617 432 1050
Grace M. Lee and W. Katherine Yih share senior authorship.
Received: June 7, 2011
Accepted: April 26, 2012
Published online: July 28, 2012
Number of Print Pages : 7
Number of Figures : 1, Number of Tables : 4, Number of References : 30
Vol. 39, No. 2, Year 2012 (Cover Date: September 2012)
Journal Editor: Feigin V.L. (Auckland)
ISSN: 0251-5350 (Print), eISSN: 1423-0208 (Online)
For additional information: http://www.karger.com/NED