Risk of Guillain–Barré syndrome after vaccination against human papillomavirus: a systematic review and meta-analysis, 1 January 2000 to 4 April 2020

Background Guillain–Barré syndrome (GBS) is a rare autoimmune disease that can follow viral infections and has in a few cases been linked to vaccinations. Pre-licensure clinical trials did not observe an association between human papillomavirus (HPV) vaccination and GBS, a post-marketing study from 2017 reported an increased relative risk. Aim We assessed the risk of GBS after HPV vaccination through a systematic literature review and meta-analysis. Methods We searched Embase, MEDLINE and Cochrane for studies reporting on the risk of GBS after HPV vaccination in individuals aged ≥ 9 years, published between 1 January 2000 and 4 April 2020, excluding studies without a comparator group. Seven studies reporting relative effect sizes were pooled using random-effects meta-analysis. We assessed quality of evidence using the GRADE approach. Study protocol was registered (PROSPERO No. #CRD42019123533). Results Of 602 identified records, we included 25 studies. Based on over 10 million reports, cases of GBS were rare. In 22 studies no increased risk was observed, while in three studies a signal of increased risk of GBS after HPV vaccination was identified. Meta-analysis yielded a pooled random-effects ratio of 1.21 (95% CI: 0.60–2.43); I2 = 72% (95% CI: 36–88). This translates to a number needed to harm of one million to be vaccinated to generate one GBS case. Quality of evidence was very low. Conclusions The absolute and relative risk of GBS after HPV vaccination is very low and lacks statistical significance. This is reassuring for the already implemented vaccination programmes and should be used in respective communication activities.

Background: Guillain-Barré syndrome (GBS) is a rare autoimmune disease that can follow viral infections and has in a few cases been linked to vaccinations. Pre-licensure clinical trials did not observe an association between human papillomavirus (HPV) vaccination and GBS, a post-marketing study from 2017 reported an increased relative risk. Aim: We assessed the risk of GBS after HPV vaccination through a systematic literature review and meta-analysis. Methods: We searched Embase, MEDLINE and Cochrane for studies reporting on the risk of GBS after HPV vaccination in individuals aged ≥ 9 years, published between 1 January 2000 and 4 April 2020, excluding studies without a comparator group. Seven studies reporting relative effect sizes were pooled using random-effects meta-analysis. We assessed quality of evidence using the GRADE approach. Study protocol was registered (PROSPERO No. #CRD42019123533). Results: Of 602 identified records, we included 25 studies. Based on over 10 million reports, cases of GBS were rare. In 22 studies no increased risk was observed, while in three studies a signal of increased risk of GBS after HPV vaccination was identified. Meta-analysis yielded a pooled random-effects ratio of 1.21 (95% CI: 0.60-2.43); I 2 = 72% (95% CI: . This translates to a number needed to harm of one million to be vaccinated to generate one GBS case. Quality of evidence was very low. Conclusions: The absolute and relative risk of GBS after HPV vaccination is very low and lacks statistical significance. This is reassuring for the already implemented vaccination programmes and should be used in respective communication activities.

Background
More than 10 years after the licensure of the first human papillomavirus (HPV) vaccines, a growing body of evidence supports the large-scale implementation of HPV immunisation programmes. Clinical trials and postmarketing observational studies have shown consistent efficacy, effectiveness and safety of the available HPV vaccines: (i) the bivalent Cervarix, (targeting HPV types 16 and 18, GlaxoSmithKline Biologicals, Rixensart, Belgium); (ii) the 4-valent recombinant Gardasil (targeting HPV types 6, 11, 16, and 18, MSD VACCINS, Lyon, France) and; (iii) the 9-valent Gardasil 9 (targeting HPV types 6,11,16,18,31,33,45,52 and 58, MSD VACCINS) [1,2]. Vaccination against HPV reduces the prevalence and incidence of cervical intraepithelial neoplasia grade 2 or 3 or worse (CIN2/3 or worse) among girls and women and anogenital warts diagnoses among girls, women, boys and men [3]. Since 2007, HPV vaccination programmes have been implemented in most European countries, usually targeting females. In recent years, several countries have extended their recommendation to a gender-neutral programme [4]. Notably, herd effects have been measured in countries with high HPV vaccination coverage [3,5].
The future public health impact of HPV vaccination on HPV-associated disease will rely on the vaccination coverage achieved. While the expansion of vaccination programmes is encouraging and uptake is increasing, overall HPV vaccination rates remain low and below national targets in a number of countries [6,7]. Suboptimal vaccination coverage is often driven by vaccine hesitancy, which in turn is often related to public debates and fear of vaccine-induced side effects [8,9].
Among the possible risks associated with vaccination, Guillain-Barré syndrome (GBS) is one of the most serious. Guillain-Barré syndrome is a rare autoimmune disease where the body's immune system damages nerve cells, causing muscle weakness and, in some cases, paralysis. Most people recover, however, some have lasting long-term weakness and GBS can be fatal. It occurs with a frequency of less than one case per 100,000 person-years in the age group relevant for HPV vaccination, i.e. those aged 10-19 years [10,11].
The causes of GBS are not yet fully understood, however, it often occurs after viral or bacterial infections and, in rare cases, after vaccination [12].
While pre-licensure clinical trials showed no association between HPV vaccination and subsequent occurrence of GBS, a French study from 2017 reported a more than threefold increased relative risk [13]. No evidence of an association between HPV vaccination and any autoimmune disorder has been found so far [9]. To the b Two studies published findings with a large overlap in data based on the VAERS dataset [34,35] with similar findings. The effect estimate by Geier et al. 2017 [35] was used in the primary analyses and Geier et al. 2015 [34] estimate in the sensitivity analysis. Therefore, the results of the meta-analysis show findings of six studies.
best of our knowledge, there is no systematic literature review investigating the potential association between the HPV vaccination and GBS specifically. We assessed the available evidence on the risk of GBS after HPV vaccination by including both randomised controlled trials (RCT) and post-marketing non-randomised studies.

Methods
We registered our systematic literature review protocol at the International Prospective Register for Systematic Reviews (PROSPERO) under the registration number CRD42019123533 [14]. We report our results according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) reporting guideline [15]. We included all studies reporting on individuals aged 9 years and older (i.e. the licenced age) who have been vaccinated against HPV with one of these vaccines: the bivalent Cervarix; (ii) the 4-valent recombinant Gardasil and; (iii) the 9-valent Gardasil 9. We included all possible vaccination schedules, including stopped schedules.
We did not restrict any study design, however, we excluded studies lacking a comparator group. and any type of control group would suffice, given the fact that there is one (e.g. placebo, no/other vaccination). We did not restrict our search based on language or geography.
The outcome of interest was GBS after HPV vaccination, as sub-defined by Brighton criteria [16] and all other non-Brighton criteria. According to the Brighton criteria, GBS includes acute inflammatory demyelinating polyradiculoneuropathy and acute motor axonal neuropathy. Other, non-Brighton criteria include Miller Fisher syndrome, which is a subtype of GBS characteristically consisting of the triad of ataxia, areflexia, ophthalmoplegia, acute motor and sensory axonal neuropathy and overlap syndromes between GBS and Miller Fisher syndrome.
We reviewed all literature reporting on the risk of occurrence of GBS after HPV vaccination, published between 1 January 2000 and 21 January 2019 and indexed in Embase, MEDLINE and the Cochrane Central Register of Controlled Trials. On 2 April 2020, we updated our search and included an additional search for publications in PubMed. We provide the full search strategy in the Supplement. Additionally, we used the snowballing approach to include additional studies by hand-searching the citation lists of included studies.

Study selection and data collection
We uploaded all records to Covidence, a screening and data extraction tool for systematic reviews. Two reviewers (TSB and TH) independently included and excluded studies, using a stepwise approach based on title and abstract screening and a subsequent full-text screening.
Subsequently, one reviewer (TSB) extracted data from the included studies using a pre-defined data extraction sheet (the data extraction sheet is included in the Supplement). The second reviewer (TH) revised the data extractions against the original papers to identify potential errors. In case of disagreement, a final decision was made by consensus between the reviewers.
From the included studies, we extracted: (i) information on the study set-up (design, location, study period and follow-up time in person-years, inclusion and exclusion criteria); (ii) study population, (sex; age; number of people included in total, and by vaccinated/control group); (iii) intervention (type of vaccine used); (iv) control group; (v) potential co-interventions and; (vi) outcome (GBS definition; source of outcome reporting; incidence in the HPV-vaccinated and control groups). When available, we also extracted the incidence rate (IR) and all reported measures of association, including the incidence rate ratio (IRR), relative risk (RR), odds ratio (OR), hazard ratio (HR) and potentially corrected confounding factors. Furthermore, we collected funding source and reported conflict of interest as risk of bias indicators.

Assessment of risk of bias and quality of the body of evidence
Two reviewers (TSB and TH) independently assessed included studies for risk of bias. For RCT we used the revised Cochrane Collaboration's tool (RoB 2.0) [17] and for non-randomised studies the Risk of Bias in Non-randomized Studies -of Interventions (ROBINS-I) tool was used [18]. The overall assessment of the quality of the body of evidence followed the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach [19] in its most recent version adapted for use of ROBINS-I [20].

Meta-analysis
All relative effect measures were pooled into one relative effect measure (ratio). Between-study heterogeneity in random-effects meta-analysis is reported through I 2 . Meta-analyses were conducted using the meta package in R version 3.6.3 (R Foundation for Statistical Computing, Vienna, Austria).
Pre-specified subgroup analyses were planned to explore the potential effect of study design, vaccine type (i.e. bivalent vaccine, 4-valent recombinant vaccine and 9-valent vaccine) and GBS case definition (Brighton vs non-Brighton) on the pooled effect estimate. When multiple studies were reporting on the same data source but with different subgroups of reporting timeframes, the most recent and/or most complete study was used for the primary analysis. Sensitivity analysis was performed by vaccine type and by outcome measurement to assess the robustness of the results of primary meta-analyses.

Table 1c
Characteristics of the included studies reporting on the risk of Guillain-Barré syndrome after human papillomavirus vaccination, 1 January 2000-4 April 2020 (n = 25)

Public health perspective
To assess the public health implications of our findings, we calculated the number needed to harm (NNH) to generate one additional case of GBS using GRADEprofiler version 3.6 (Informer Technologies, Los Angeles, United States (US)), based on the pooled findings of this study and the baseline risk for the age group of 10-19-year-olds (males: 0.97/100,000; females: 0.55/100,000; overall 0.75/100,000) [21]. For comparison, the number needed to vaccinate (NNV) was calculated in respect to the prevention of one case of cervical cancer [22].

Results
In total, we identified and screened 602 citations and included 25 studies (Figure 1).
While most study populations comprised of adolescents, older and younger participants were also included. Age ranged from 6-72 years. We planned for inclusion of people aged 9 years and older [14] because the HPV vaccines are approved for people aged 9 years and older. However, for completeness, we decided not to exclude studies reporting on a broader age group. The majority of studies reported exclusively on the vaccination of girls and women. Twelve studies reported exclusively on females (100% females in 11 studies and 97% females in one study), nine studies reported on both sexes (some of which had males only as comparator), one study reported exclusively on males and three studies did not report the sex of the participants ( Table  1).
The included studies comprised data of more than 10 million reports in total. Of 25 studies, 14 reported the number of cases/controls, reports or vaccine doses (range: 4,133,370-4,415,894), 10 studies reported the number of participants (range: 6,622,607-6,843,326) and one study [23] did not report any number of participants, reports or case/control numbers.

Comparator
Studies compared the risk of occurrence of GBS after HPV vaccination to either the risk of GBS after no vaccine (e.g. non-targeted populations such as boys), to another vaccine e.g. hepatitis B (HBV vaccine), meningococcus or influenza vaccine), or against historical background rates ( Table 2). Case-control studies compared frequency of HPV vaccination in GBS cases to the frequency of HPV vaccination in controls; the controls being the general population or other adverse events reported to the vaccine safety registry.

Outcome
The duration of follow-up time varied, ranging from 42 to 180 days post-vaccination to the total accumulated number of person-years available in the registry or medical records. Eight studies reported the total follow-up time [13,[26][27][28][29]33,36,37], which adds up to 42,055,425 reported person-years in total.
While five studies referred to the Brighton criteria for GBS case definitions [30,38,39,41,44], in the majority of studies GBS diagnosis was based on original and expanded International Classification of Diseases codes (ICD-9 and ICD-10) as well as other coding systems such as Medcode, MedDRA and VAERS and freetext notes in medical files ( Table 2).

Risk of bias
We summarised the findings of our risk of bias assessment in Figure 2 and Supplementary Table S3. The risk of bias in the community-based RCT by Lehtinen et al. and Bi et al. [24,25] was considered to be low for most indicators, but high regarding the selection of the reported results, which was limited for GBS. We were not able to assess the risk of bias in the pooled analysis of 42 studies by Verstraeten et al. [46], of which 16 reported HPV vaccination, because of the complexity of the pooled design and lack of reporting of the key indicators for the risk of bias assessment. The risk of bias in 14 of 22 non-randomised studies was assessed as being critical and eight studies were assessed as being at serious risk of bias. The risk of bias was mostly introduced by the critical risk of confounding (i.e. lack of confounding correction), or because of the outcome measurement i.e. GBS diagnosis not based on Brighton criteria [15]. Risk of bias because of the classification of the intervention (vaccination status) was moderate when based on a registry and critical when self-reported. Less than 10 studies per outcome and per study design were available for pooling, which  2016 [27] Register-based open cohort

Table 2d
Occurrence of Guillain-Barré syndrome and association between Guillain-Barré syndrome and vaccination status, 1 January 2000-4 April 2020 (n = 25) prevented us from systematically assessing publication bias by means of a funnel plot.

Results of individual studies
The reported occurrence of GBS following HPV vaccination is summarised in Table 2 and described by study design and by geographic region.

Randomised studies
Two randomised studies did not observe a single GBS case among people who received the HPV vaccine and found no increased risk of GBS after HPV vaccination. Lehtinen et al. [24] and Bi et al. [25] reported on a large community-based RCT in Finland. Among 32,176 adolescents, no GBS cases were diagnosed after the bivalent vaccine (or HBV vaccination). Verstraeten et al. [46] described the findings of a pooled analysis of all RCT of AS04-adjuvanted bivalent vaccines, HSV and HBV vaccines. One GBS case was observed among 68,512 participants, in the control group.

Non-randomised studies
One cohort study [13] and two case-control studies [30,31] investigated the potential association between the HPV vaccine and autoimmune disease in France.
In the cohort study of more than two million girls by Miranda et al. [13], an increased risk of GBS was observed among vaccinated girls. There was an IR of 1.36 cases per 100,000 person-years among vaccinated individuals (20 cases), compared with 0.37 cases per 100,000 person-years among unvaccinated individuals (23 cases), with an adjusted hazard ratio (aHR) of 3.78 (95% CI: 1.79-7.98). The association was particularly marked in the first 2 months after vaccination and decreased over time, and did not differ with the type of HPV vaccine or whether or not GBS was preceded by a recent history of gastrointestinal or respiratory tract infection. Seasonality and calendar year did not affect the findings. In the two included case-control studies, no exposure to HPV vaccine was observed in cases with GBS [30,31]. In Finland, Skufca et al. [26] investigated baseline annual, pre-vaccination, and post-vaccination period incidences of GBS, followed by a nationwide population-based observational cohort study among 240,605 women that compared the risk of GBS between those exposed and not exposed to the bivalent vaccine. There was no increase in GBS incidence in the post-vaccination period in neither men nor women. Cox regression showed a statistically non-significant trend towards increased risk of GBS after HPV vaccination among women, with an aHR of 5.
Additional subgroup analysis by Brighton vs non-Brighton GBS case definition was not possible because all seven studies used non-Brighton criteria. The included studies were heterogeneous in terms of study design, analysis and reporting, which was also reflected in the high I 2 , which indicates the percentage of variability in the effect sizes which is not caused by sampling error.

Quality of the body of evidence
According to the GRADE approach, the quality of the body of evidence is very low. We had to downgrade the quality of evidence three times: (i) for risk of bias (residual confounding at least); (ii) for imprecision (wide 95% CI around the pooled estimate including substantial benefit as well as substantial harm) and; (iii) heterogeneity (inconsistency; I 2 = 72%).

Number needed to harm and number needed to vaccinate
Translating the pooled random-effect ratio of 1.21 (95% CI: 0.60-2.43) to the number needed to harm (NNH), we estimated that one million people need to take the HPV vaccine to generate one case of GBS (95% CI: − 3 to 8 cases). In contrast, 324 (80% credibility interval: 195-757) people need to be vaccinated to prevent one cases of cervical cancer [48].

Discussion
The results of our systematic review and meta-analysis indicate that absolute risk of GBS after HPV vaccination is low. Reported historical background rates of GBS incidence were between 0.55 and 2.25 cases per 100,000 person-years [38,39]. A slightly increased RR of GBS after HPV vaccination is low, far away from statistical significance based on findings from our metaanalysis. From a public health point of view, up to one million people would need to be vaccinated to generate one additional case of GBS, while the NNV to prevent one case of cervical cancer is ca 300.
We performed a comprehensive and systematic search on this topic that includes all licenced HPV vaccines. It was specifically targeted at investigating the association with GBS, in the context of other reviews that focussed on the potential association between HPV vaccination and autoimmune and/or neurological diseases [49][50][51].

Figure 3
Meta-analysis of studies reporting an effect estimate of the risk of Guillain-Barré syndrome after HPV vaccination, by study design, 1 January 2000-4 April 2020 (n = 7) a,b
The quality of our findings depended on the quality of the studies, which were largely registry studies and based on non-Brighton GBS outcomes. There was risk of bias because of large heterogeneity in the design and reporting of the studies, as well as the control groups. Confounding was the biggest limiting factor of the quality of the evidence, because many studies were not designed to correct for confounding and cohorts were highly confounded. Often, the control group was not matched based on sex or age group. Outcome ascertainment was challenging, given the heterogeneity of case definitions for GBS; only one of five studies used the Brighton criteria.
The follow-up period for the detection of GBS varied between studies and many of the cohorts or registrybased studies were partially underpowered by design because of the rarity of GBS. In the three studies that signalled increased risk of GBS after HPV vaccination, conflicting temporal trends were reported. In the study by Miranda et al. [13] the association between vaccination and GBS was particularly marked in the first 2 months after vaccination and decreased over time, while Skufca et al. [26] reported a substantial increase in the association (with very wide CI). Souayah et al.
[29] also noted increased reporting of GBS during the first 6 weeks after vaccination, although interpreting this trend was challenging in this VAERS reporting study since all five other studies reporting VAERS data found no association between HPV vaccination and GBS. Interpretation of these results in terms of causality should be made with caution.
In future studies, consensus on the case definition for GBS and the risk timeframe is needed to generate uniform and comparable findings. Studies in settings with gender-neutral vaccination policies are required to further assess the risk of GBS after HPV vaccination among boys and men. Furthermore, studies should be expanded to geographical areas outside of western Europe and North America, where HPV vaccination is being implemented on large scale. To generate the highest quality of evidence on this topic, we recommend further research with a self-controlled case series design using Brighton-outcomes. The self-controlled case series design has been proved most suitable for rare events and limits confounding to time-dependent confounding [21], as also used to assess the risk of intussusception after rotavirus vaccination [52].
This study aims to provide up to date vaccination safety information for healthcare providers and policymakers as well as the general public [53]. Transparent communication of potential safety issues is essential to build trust and strengthen confidence in HPV vaccination. Concern about vaccine safety is one of the key determinants of vaccine hesitancy and poses a threat to public health. Healthcare providers play an important role in communicating information on HPV safety [54]. The low potential risk of GBS after HPV vaccination should have minimal impact on the risk consideration for HPV vaccination programmes, reassure vaccine confidence and ultimately increase vaccination rates.