Is Oseltamivir (Tamiflu®) Safe? Re-examining the Tamiflu 'Ado' from Japan

Etsuji Okamoto


Expert Rev Pharmacoeconomics Outcomes Res. 2010;10(1):17-24. 

In This Article

Surveys/Studies in Japan

Since complications such as encephalopathy related to influenza attracted a lot of attention, research projects were instituted to conduct epidemiological surveys with grants from the MHLW.

Yokota Project

The first project on the complications of influenza was instituted in 2005 with Yokota as a principal investigator.[104] The project distributed survey forms to pediatricians in 12 prefectures and asked them to report ten consecutive pediatric patients with laboratory-confirmed diagnoses of influenza in the 2005/2006 season and to provide survey forms to family caregivers of the surveyed patients. They collected 2846 survey forms on patients and 2545 survey forms from family caregivers. The incidence of abnormal behaviors was 11.9% in patients prescribed oseltamivir and 10.6% in patients otherwise, the difference of which was not statistically significant.

Detailed statistical analysis was reported separately by Fujita, a coinvestigator.[9] He compared the incidence of NPAEs between oseltamivir and acetaminophen. He demonstrated an elevated incidence of abnormal behaviors with acetaminophen, with a hazard ratio of 1.30 (p = 0.079), adjusted for risk factors, including fever, by multivariate analysis of the proportional hazard model. By contrast, no NPAEs were significantly increased with oseltamivir.

Hirota Project

The Yokota project repeated the survey in the 2006/2007 season using the same methodology but with an expanded scale. They distributed survey forms to 697 hospitals/clinics and collected 10,745 patients data (10,316 forms filled in by doctors and 10,103 forms filled in by family caregivers with 9674 overlapping) during the period of 11 January to 20 March 2007. However, the project was disbanded because of a conflict-of-interest issue and the data were succeeded by a new project led by Hirota.[105]

The Hirota project analyzed 9666 patients after excluding those aged 18 years or over (n = 21), those who had developed abnormal behaviors before doctors' visits (n = 351) and those whose data on abnormal behaviors were missing (n = 278).[105] The incidence of abnormal behaviors was lower in those prescribed oseltamivir (840 out of 7438; 11%) than those not prescribed oseltamivir (286 out of 2228; 13%) with a p-value of 0.046. However, their conclusion was inconclusive: "since abnormal behaviors could originate from influenza per se, it was difficult to draw any firm conclusion from the given data alone".[10]

Yorifuji et al. challenged the methodology adopted by the Hirota project, claiming that their reanalysis using person–time data had yielded a significantly elevated risk of abnormal behaviors with oseltamivir. The data reported in Yorifuji's article are slightly different from the final report of Hirota's project. This was because Yorifuji used the data from the interim report presented earlier.[10]

The Yokota team moved 99 patients whose adverse behaviors had preceded the treatment with oseltamivir from the prescribed group to the nonprescribed group. Hama criticized this approach from the viewpoint of intention-to-treat (ITT) analysis.[106] To this claim, Hirota defended in his report that "ITT is the method for randomized controlled trials and not applicable to observational studies". A dictionary of epidemiology defines ITT as "a fundamental way to analyze subjects in an randomized controlled trial",[11] lending support to Hirota's defence.

Solicited Reporting of NPAEs

Pursuant to the Pharmaceutical Affairs Act, doctors/pharmacists are required to report any adverse events if they consider them to be caused by drugs. However, they are not required to do so just because patients showed any adverse events in the course of illnesses. Pharmaceutical manufacturers are required to report ADRs only when they are related to their products. There is no reporting system to capture all adverse events occurring in the usual course of illnesses.

The National Institute of Infectious Diseases (NIID) started a 'solicited' reporting of serious NPAEs observed in influenza patients to estimate the incidence. The government circulated a letter asking for collaboration to all clinics/hospitals of internal medicine or pediatrics in the country (~80,000). Serious NPAEs were defined as 'sudden running', 'jumping out' or 'any unexpected behaviors which could be life-threatening if they are not stopped'.

The survey was started in the 2007/2008 season and collected cases retrospectively in the 2006/2007 season and prospectively in the 2007/2008 season and after. Reporting was collected by either internet or fax and data on body temperature, medication and timing were collected. Throughout three seasons, a total of 200 serious NPAEs were reported, of whom 71 (35.5%) were prescribed oseltamivir (Table 1). NIID also estimated the incidence by dividing the reported cases with the number of patients estimated by their sentinel surveillance and concluded that there were no significant difference in the incidence of NPAEs in teenagers before and after the warning against use of oseltamivir to teenagers on 20 March 2007.

Findings of the spontaneous reporting and the NIID survey are summarized in Table 1, supplemented with the severity of epidemic in each season (expressed as the average number of cases per sentinel compiled by the NIID) and percentage of oseltamivir in prescription compiled from the Japan Medical Data Center, Inc. (JMDC) database will be explained later.

Case-series Studies

Clinicians, as well as activists, reported their own findings in independent studies/surveys in academic journals. Although their strength of evidence is limited owing to their study design, these publications provide some insights.

Hama reported five fatal and three nonfatal NPAEs related to oseltamivir.[12] All fatalities overlap with sudden deaths cases reported by Chugai pharmaceuticals but the author supplemented the description by interviews with the surviving family and reviewing their medical records. The author claims that sudden deaths are major adverse events of oseltamivir. This theory may be justified in view of the fact that sudden deaths are not observed in other drugs (only one accidental death was reported for zanamivir in the 2008/2009 season).

Three case-series reports were reported by clinicians of a community hospital in Osaka. Hara and others reviewed 18 pediatric influenza patients who had showed NPAEs in 1219 pediatric influenza patients in 2005/2006 season and concluded that "abnormal behaviors observed in 18 patients were considered as febrile delirium".[13] Tominaga and others identified 26 pediatric influenza patients who had showed NPAEs out of 240 pediatric patients who had visited the emergency room of their hospital in 2006/2007 season.[14] Of these patients, 18 were diagnosed with influenza and 13 of them had taken oseltamivir. The author concluded that "prominent motor symptoms, recurrence of abnormal behaviors were less frequent with Tamiflu administration and it was suggested that Tamiflu neither worsened nor promoted the abnormal behaviors". Tanabe and others reported 22 pediatric influenza patients (mean ± SD: 6.6 ± 3.2) admitted to their hospital in three seasons (2004/2005, 2005/2006 and 2006/2007) who had been treated with oseltamivir and demonstrated abnormal behaviors. They compared 13 patients who had developed abnormal behaviors before oseltamivir treatment and nine patients who had developed abnormal behaviors after oseltamivir treatment and concluded that "oseltamivir is not a prerequisite for the development of abnormal behaviors and does not worsen them".[15]

Goto and others reviewed 50 hospitalized pediatric influenza patients in 2005/2006 season. Abnormal behaviors constituted the largest share of the reason for admission (14 out of 50 or 28%), including a 12.5-year-old boy who had jumped out of the window. The boy developed 39.0 centigrade fever at 19:00 and ingested oseltamivir at 20:00. He jumped out the window at 23:00 but, fortunately, remained uninjured. He took another pill of oseltamivir at 14:00 on the next day but did not develop any abnormal behaviors. The boy remembered his behaviors well but was not able to explain why he had done so. The author dismissed the causality between oseltamivir and abnormal behaviors of the boy because the abnormal behaviors were not reproducable after the repeated ingestion of oseltamivir.[16]

Tomono and others conducted a cross-sectional study on the pediatric influenza patients treated at their hospital in 2006/2007 season. A total of 12 cases who had visited their hospital for NPAEs (boys:girls: 9:3; age: 8.25 ± 3.22 years) were matched with 335 controls (183 out of 152; 6.09 ± 3.74). Six cases (50%) in the case group had taken oseltamivir prior to NPAEs in contrast to 261 controls (77.9%), the difference of which was statistically significant (p = 0.024) and the authors concluded that oseltamivir had significantly reduced the incidence of NPAEs in patients with influenza.[17]

Although these case-series are of limited evidence owing to their study design, they do provide some insights on the clinical symptoms of pediatric influenza patients. Abnormal behaviors including potentially dangerous ones such as jumping out are not so 'abnormal' events in daily practices of pediatricians. In fact, as Goto pointed out, abnormal behaviors are the most common reasons for admission. This will explain why most pediatricians do not consider abnormal behaviors observed among pediatric influenza patients as isolated drug adverse events.

Incidence of Injuries using Health Insurance Claims

I have surveyed the incidence of injuries shortly after contracting influenza based on the belief that if suspected abnormal behaviors are true, the danger will be the injuries and subsequent deaths, not the abnormal behaviors themselves. If influenza patients sustained injuries, they are likely to visit surgery clinics or emergency department, and not the prescribing pediatricians. Surveys on pediatricians alone may not be able to detect injuries. Claims data may better be suited for linking influenza and injuries because all claims regardless of providers will be submitted to the health insurer. The author analyzed the incidence of injuries within 3 days of the diagnosis of influenza by using health insurance claims to determine if oseltamivir is related with increased incidence of injuries.[18]

Health insurance claims of seven health insurance societies with the cumulative enrollment in five influenza seasons of approximately 1 million were analyzed. By law, clinics and pharmacies are required to submit claims following a uniform format, a uniform fee schedule and a uniform formulary. Information contained in health insurance claims submitted from clinics and pharmacies for reimbursement were submitted into a database managed by JMDC on a contractual agreement. The input data are encrypted with a unique patient identifier, patient characteristics such as age, sex, diagnoses contained in medical claims (coded into ICD10), as well as diagnosed dates, generic drug codes and dispensing date contained in pharmaceutical claims. Diagnoses within 6 months of the diagnosis date of influenza were also collected to be used as covariates.

Medical claims containing diagnoses of influenza (ICD10 code: J111) with diagnosing dates within the five influenza seasons (December to March of the years 2003/2004–2007/2008) were extracted and matched with pharmaceutical claims as well as other medical claims from different clinics or hospitals such as surgery clinics using unique patient identifiers. Diagnoses of injury (ICD10 S category) as well as their date of diagnoses were also extracted from medical claims. To capture injuries sustained shortly after the ingestion of the target drugs, only injuries diagnosed within 3 days of the diagnosis of influenza were extracted.

A total of 140,800 influenza cases were identified in five influenza seasons, allowing for duplicates of the same individual over seasons, of which 51,505 (36.6%) were prescribed oseltamivir (the percentage of those prescribed oseltamivir is much lower than 90% of the Yokota report and 77% of the Hirota report). This discrepancy is explained by the difference in target population. Both Yokota and Hirota surveyed on pediatricians, while my data include all ages. Since oseltamivir was available in dry syrup and quick on-site test kits were widely available among pediatricians, it was common among Japanese pediatricians to prescribe oseltamivir until the warning was issued by MHLW in 2007). A total of 97 injuries were diagnosed within 3 days of diagnoses of influenza (6.9/10,000), of whom 24 were prescribed oseltamivir. The incidence was 4.7 for those prescribed oseltamivir and 8.2 for those otherwise per 10,000, the difference of which was statistically significant (p = 0.015). The observed tendency that those prescribed oseltamivir had lower incidence of injuries than those otherwise was consistent over five subgroups of similar propensity score, although none of them reached a statistical significance owing to smaller sample size (Table 2). A logistic regression analysis with the injuries as a dependent variable revealed that Tamiflu was a covariate with a negative effect on the incidence of injuries (OR: 0.622; 95% CI: 0.384–1.007).


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.