A Nationwide Epidemiological Survey of Adolescent Patients With Diverse Symptoms Similar to Those Following Human Papillomavirus Vaccination: Background Prevalence and Incidence for Considering Vaccine Safety in Japan

Background Since June 2013, Japan has suspended proactive recommendation of human papillomavirus (HPV) vaccination due to self-reported diverse symptoms, including pain and motor dysfunction, as possible serious adverse events following immunization. Although these symptoms may be seen in adolescents without HPV vaccination, their frequency, taking into account disease severity, has not been examined. Methods A two-stage, descriptive, nationwide epidemiological survey was conducted in 2016, with a 6-month target period from July 1 to December 31, 2015, to estimate the prevalence and incidence of diverse symptoms among Japanese adolescents without HPV vaccination. Participants were 11,037 medical departments in hospitals selected nationwide by stratified random sampling. Eligible patients had to satisfy four criteria: (1) aged 12–18 years upon visiting hospital; (2) having at least one of four symptoms/disorders (pain or sensory dysfunction, motor dysfunction, autonomic dysfunction, or cognitive impairment); (3) symptoms/disorders persisting for at least 3 months; and (4) both criteria (2) and (3) influence attendance at school or work. We then extracted data of patients with diverse symptoms similar to those after HPV vaccination while considering opinions of doctors in charge. Results Estimated 6-month period prevalence of diverse symptoms among girls aged 12–18 years without HPV vaccination was 20.2 per 100,000. Annual incidence was estimated to be 7.3 per 100,000. Conclusion Adolescent Japanese girls without HPV vaccination also visited hospitals with diverse symptoms similar to those following HPV vaccination. Our findings predict the medical demands for coincident diverse symptoms, which are temporally associated with but not caused by HPV vaccination of Japanese adolescents.

persisting for at least three months; and (4) both criteria (2) and (3) influence attendance at 107 school or work. These criteria were applied uniformly regardless of the department that was 108 invited for the survey. Survey explanation forms are shown in Supplementary Material 1. If 109 departments had one or more patients, they were asked to report the number of patients by 110 gender and age (1-year intervals). The numbers of eligible boys and girls were reported. As of 111 2015, that is, at the time of this survey development, gender neutral HPV vaccination had 112 already been recommended in the USA, Canada, Austria, and Australia. 19 We thought it would 113 be meaningful to evaluate the background frequency of diverse symptoms for both genders in 114 Japan. 115 If departments that responded in the first-stage survey had one or more patients that satisfied 116 the criteria, the doctors in charge were invited to the second-stage survey to provide the 117 following clinical characteristics for each patient based on pre-exiting medical charts: previous 118 medical history, date of/age at symptom onset, date of the first/last visit, history of HPV 119 vaccination, presence/absence of each symptom (during July to December 2015), duration of 120 the symptoms, status of schooling/working attendance (during July to December 2015), and 121 diagnoses that were identified/recognized by departments (up to 10 diagnoses, regardless of the 122 cause for the present symptoms, but not providing any diagnoses where examinations were only 123 requested). We also solicited the opinions of doctors in charge about whether they were able to 124 adequately explain the patient's symptoms during July to December 2015 by their diagnoses 125 (no/yes/unknown), and, if yes, which diagnosis was the most explicable. Informed consent was 126 waived because anonymity was maintained during information collection. The study protocol 127 was approved by the ethics committee at Osaka University, Graduate School of Medicine (No. process 132 At the time of information collection, symptoms of the reported patients did not necessarily 133 correspond to those after HPV vaccination. Thus, we further defined whether those symptoms 134 corresponded to "diverse symptoms after HPV vaccination" based on two kinds of information 135 in the second-stage survey: "opinions of doctors in charge" whether or not they were able to 136 adequately explain patients' symptoms by the diagnoses that were identified/recognized by the 137 department, and "a diagnosis" that was designated as the most explicable by the doctor in charge 138 ( Table 2). Our decision harmonized with a claim in Japan that "there have been no diverse 139 symptoms before the introduction of HPV vaccine", and a consensus of the National Expert   147 Using information from the first and second stage survey, we calculated the period prevalence 148 (from July 2015 to December 2015) of patients aged 12-18 years with diverse symptoms by 149 gender and HPV vaccination history ( Figure 1). We limited our estimation to patients whose 150 age at onset was ≥12 years because HPV vaccine is designated as NIP for girls aged 12-16 years 151 in Japan. HPV vaccination history was classified into five categories: "never vaccinated (Group 152 A)"; "vaccinated, symptom occurred before vaccination (Group B)"; "vaccinated, symptom 153 occurred after vaccination (Group C)"; "vaccinated, the time relation between vaccination and 154 symptom onset was unknown (Group D)"; and "unknown vaccination status (Group E)". All 155 boys were considered "never vaccinated (Group A)", since HPV vaccine is not designated as

173
Of the 18,302 departments selected for the first-stage survey, 11,037 (60.3%) responded (Table   174 1 and Supplementary Table 6). A total of 508 departments reported one or more patients who   178 The estimated number of patients aged 12-18 years with diverse symptoms, whose age at onset 179 was ≥12 years during the period of July to December 2015, was 829 boys and 1,590 girls (Table   180 3). The corresponding 6-month period prevalence of diverse symptoms was 19.8 per 100,000   189 Of 508 departments that reported one or more patients who satisfied criteria in the first-stage   201 We calculated period prevalence of diverse symptoms according to the number of symptoms 202 among girls separately for Group A and Group C ( Figure 2) by using period prevalence (Table   This was the first nationwide survey to estimate prevalence of severe diverse symptoms among 222 adolescents without a history of HPV vaccination in Japan. We confirmed that adolescents with 223 no history of HPV vaccination also visited hospitals due to diverse symptoms and a wide 224 spectrum of diagnoses, demonstrating that diverse symptoms include elusive and heterogeneous 225 disease entities. We also estimated the annual incidence of diverse symptoms among 226 unvaccinated girls, which predicts the medical demand for diverse symptoms that 227 coincidentally occur following vaccination of Japanese female adolescents. A response rate 228 greater than 60% in each stage of the survey would ensure a certain degree of generalizability.

229
There may be some criticism that our survey did not accurately focus on unvaccinated girls 230 with diverse symptoms in "a multilayered way", since one eligibility criteria for reported patients was "having at least one of the symptoms". However, even girls with a single symptom 232 have been reported to the MHLW as "not recovered" cases of adverse events following HPV 233 vaccination. 15 We also found that both 6-month period prevalence and annual incidence among 234 unvaccinated girls with ≥10 symptoms were "not zero" when number of symptoms was 235 considered as a surrogate variable of multilayered symptoms.

236
In addition to the prevalence of diverse symptoms among girls without a history of HPV 237 vaccination (Group A), we estimated the prevalence among vaccinated girls whose symptoms 238 occurred after vaccination (Group C). However, these estimates cannot be directly compared 239 between groups because suspension of the proactive recommendation for HPV vaccination in 240 Japan led to a smaller vaccinated population among girls aged 12-14 years (Table 5). Some 241 unstable age-specific prevalence due to a sparse denominator (e.g., for aged 14 years in Group 242 C) prevented us from obtaining reliable age-standardized estimates. Furthermore, there were 243 many potential biases that could have led to underestimation or overestimation of prevalence 244 in each group (Table 6). Nonetheless, it is noteworthy that the prevalence in Group A seemed 245 to be underestimated overall because situation No. 1 in Table 6 (i.e., if patients already have 246 diverse symptoms, they will not receive HPV vaccine), which results in overestimation of 247 prevalence, would be very unlikely due to low HPV vaccination coverage at the time of the 248 survey.

249
There are other reasons why we cannot compare prevalence between unvaccinated and 250 vaccinated girls. First, there were 604 girls with diverse symptoms in Group E whose 251 vaccination status was unknown (Table 5) Figure 2). Second, patients in Group C included girls whose diverse 257 symptoms occurred more than several months after last HPV vaccination (maximum: 48 258 months in the second-stage survey). When we limited Group C to those who had shorter 259 durations, the period prevalence was lower than that in Group A (Supplementary Table 7). The 260 proportion/number of each symptom between groups in the second-stage survey 261 (Supplementary Figure 2) also cannot be compared due to potential bias, as shown in Table 6.

262
Despite these biases, we believe that providing proportions/numbers, with careful interpretation, 263 is meaningful for a better understanding of HPV vaccine issues in Japan.

264
Reports similar to ours have not yet been published in terms of severity of diverse symptoms.

265
The previous study in Nagoya, Japan, which targeted diverse symptoms among Japanese 266 adolescent girls, differed from our survey not only because it compared the odds of various 267 symptoms in vaccinated and non-vaccinated girls, but also because it evaluated milder 268 symptoms: a cumulative incidence in unvaccinated girls of 0.2 to 25.6%. 14 While that study 269 showed no association between HPV vaccination and 24 symptoms by comparing vaccinated 270 and unvaccinated girls, another study re-analyzed the data and showed possible associations for 271 several symptoms, such as cognitive impairment and movement disorders. 23   vaccines (e.g., hepatitis A virus vaccine) as controls. 24-34 Overall, the safety profile was similar 284 between the HPV vaccine and control groups. These reports also consistently showed that the 285 most frequently reported solicited adverse reactions following vaccination were injection site 286 symptoms (pain, redness, and swelling), which were reported significantly more often in the 287 HPV vaccine group than in the control group, but were generally transient.

288
Orthostatic dysregulation and its associated disorders were the most frequent diagnoses, and 289 could not be clearly distinguished from diverse symptoms after HPV vaccination. Our findings 290 were in line with the fact that POTS has been frequently evaluated as a possible adverse event 291 following HPV vaccination, 35-39 although the majority of reports found no significant safety 292 concern or supportive evidence for a causal relationship. The annual incidence of POTS among 293 unvaccinated girls was estimated to be 6.49 per 100,000 in Finland, 40 which is consistent with 294 our findings in terms of disease rarity.

295
The underlying mechanisms of diverse symptoms that can occur among unvaccinated girls are Some limitations in our survey should be mentioned. First, our sampling frame included all 317 hospitals in Japan, but did not include clinics because the number of clinics was tenfold or more 318 than that of hospitals. Although severe or difficult to diagnose cases are referred to hospitals, 319 and patients can visit hospitals directly without being referred by clinics in Japan, our 320 prevalence results among unvaccinated girls remained underestimated. Second, we could not 321 analyze the frequency of diverse symptoms by HPV vaccine types among vaccinated girls 322 because we did not obtain such information during the survey. Third, we could not estimate 323 confidence interval (CI) of prevalence. The protocol for a nationwide epidemiological survey 324 on intractable diseases provided a formula to calculate 95% CIs for prevalence. [16][17][18] However, 325 the formula relied on data from the first-stage survey. Since our prevalence was calculated with 326 many parameters from the second-stage survey, estimation of CIs was technically challenging.

327
Even if possible, epidemiological studies are required to consider the influence of chance, bias, 328 and confounding in a balanced manner. Since our survey was subject to substantial bias in 329 comparing prevalence of diverse symptoms between unvaccinated and vaccinated girls, the 330 highest priority was appropriate interpretation of prevalence of diverse symptoms among unvaccinated girls. This priority led us to not attempt estimation of CIs, as well as statistical 332 testing, because evaluating the extent of chance would not make sense under the considerable 333 influence of bias.

334
In conclusion, our nationwide epidemiological survey in Japan revealed that adolescent girls 335 without history of HPV vaccination also visited hospitals with diverse symptoms similar to 336 those following HPV vaccination. We hope our results help explain the complexity of HPV 337 vaccine safety issues in Japan, and will provide useful insight into future strategies of signal 338 assessment immediately after detection of AEFI.           c Denominator of the proportion was 594 patients who were reported to the second-stage survey and whose age at onset was ≥12 years (see Supplementary Figure 1 and

Supplementary Tables 1 to 5).
A c c e p t e d V e r s i o n   A c c e p t e d V e r s i o n A c c e p t e d V e r s i o n