Bulk materials were produced at large scale using embryonated hens’ eggs at the licensed manufacturing facility used for the production of the seasonal vaccine and adapted for production of the avian strain in accordance with the WHO’s biosafety guidelines for the production and quality control of pandemic human influenza vaccines. ¹² Master and working seed viruses were prepared from the NIBRG‐14 virus using specific pathogen‐free embryonated hens’ eggs. The seed virus was grown to high titer in the allantoic cavity of embryonated hens’ eggs prior to harvesting. Harvested virus was clarified, concentrated and purified using filters and ultracentrifugation and then inactivated with formalin and filtered through a 0·2 μm membrane filter for bulk material preparation. The vaccine was mixed with 0·3 mg/ml aluminum hydroxide and 8 μg/ml thimerosal (as a preservative). One‐ or ten‐milliliter vials were filled with 15 μg HA/dose and stored at 6–10°C until use. The aluminum hydroxide gel was carefully emulsified with the viral particles since stirring was required to prevent the sedimentation of inactivated whole virus adsorbed with aluminum hydroxide and to maintain the consistency and concentration of active ingredients in every vial during filling. Validation was also required for this process. The HA content of the bulk vaccine was determined by means of SDS‐PAGE (sodium dodecyl sulfate – polyacrylamide gel electrophoresis) of the purified virus before inactivation as single radial immunodiffusion (SRID) tests were unavailable.

The ovalbumin and endotoxin content were measured as indicators of impurity. These contaminants were removed using sucrose density gradient ultracentrifugation and ultrafiltration techniques.

In the formulation described here, the ovalbumin concentrations were below 1·2 ng/ml (0·6 ng/single human dose), which is significantly lower than the WHO recommended allowance, i.e. not more than 5 μg/single human dose. ¹³ In addition, bacterial endotoxin in the bulk material was not more than 7·46 EU/ml (3·73 EU/single human dose), which was also significantly less than European requirements, i.e. not more than 100 EU/dose. ¹⁴ A critical control measure for maintaining reduced endotoxin levels was close inspection of the health practices in place at the manufacturing facility where the eggs were being collected for study purposes, including sanitation practices and chicken vaccination status.

Viruses used for vaccine development were inactivated by formalin. However, as inactivation efficacy can be affected by the viral and formalin concentrations and the inactivation temperature, validation procedures must be in place before large‐scale vaccine production occurs to insure product consistency.

Inactivation efficacy and consistency were determined by passaging formalin‐treated virus samples through embryonated eggs. Inactivation was confirmed if hemagglutination activity was negative after three passages.

Previously, we produced bulk vaccine formulations derived from the Vietnam (NIBRG‐14) strain (clade 1), the Indonesia/5/05 strain (clade 2‐1) and the Anhui/1/05 (clade 2‐3) strain; ¹⁰ however, modification of each strain by RG manipulations resulted in strains with increased growth rates but differing antigen yields. As part of the success of any pre‐pandemic formulation is product availability, utilization of RG methodologies may have the potential of providing a vaccine virus that retains its antigenicity while giving good growth, thereby providing sufficient vaccine material.

Aluminum hydroxide gel made from aluminum potassium sulfate and sodium phosphate (Alum) was selected as the adjuvant for the influenza vaccine described. We have established 10 different control tests for the validation of aluminum hydroxide efficacy (Table 1). Five of these tests were derived from the Minimum Requirements for Biological Products Guidelines of Japan (hereafter Japanese Minimum Requirements or JMR), ¹⁵ one of these tests was derived from the Japanese pharmacopoeia, ¹⁶ and the other four tests were established by our laboratory (BIKEN). Criteria for the tests (e.g. tests for aluminum content and protein adsorption) were established by trend analysis and are important to maintain the lot‐to‐lot consistency on the quality of aluminum hydroxide gel.

Before aluminum hydroxide was used as an adjuvant, its capacity to absorb whole virus particles had to be determined. We demonstrated that 0·3 mg/ml of aluminum hydroxide gel had the capacity of absorbing more than 15 μg HA/0·5 ml, which was the dose selected for this vaccine formulation.

We have established two kinds of tests for defining the HA content (2, 3): SRID and calculating the HA content of bulk material by defining the ratio of purified virus suspension (before inactivation) to the bulk protein content by SDS‐PAGE. In most cases, we were unable to obtain SRID data in a timely manner, therefore we determined the HA content by SDS‐PAGE.

We have compared the results of SDS‐PAGE and SRID testing between Japanese manufacturers and the NIID and found inconsistent results in SDS‐PAGE. Therefore, harmonization of SDS‐PAGE analysis for the HA content ratio being carried out was recommended. In addition, quick preparation of reagents and determination of the standard antigen titers for SRID by the National Control Laboratories and/or the WHO collaborating centers are needed.

The process of assessing the HA content of adsorbed vaccines could not be determined using conventional SRID; however in our laboratory, we have measured HA content using a novel method developed by Kaketsuken (Kumamoto, Japan). This method consisted of eluting viral HA adsorbed to aluminum hydroxide with 0·5 m citrate–phosphate buffer (pH 6·5) and then examining using SRID. Specifically, 140 μl of citrate–phosphate buffer and 60 μl of 10% Zwittergent (CALBIOCHEM, La Jolla, CA, USA) were added to 400 μl of inactivated adjuvanted whole‐virion vaccine containing aluminum phosphate and incubated in a water bath at 37°C for 2 hours with agitation (10 seconds) every 30 minutes. The mixture was then centrifuged at 2500 × g for 5 minutes at 4°C and the supernatant carefully applied to a standard SRID gel after making appropriate dilutions with phosphate‐buffered saline.

The presence of pyrogenic substances was controlled by two different tests: bacterial endotoxin tests and the pyrogen test using rabbits. ¹⁵ Our bulk materials were pyrogen‐free, and heat‐inactivation for heat labile pyrogenic factors was not necessary.

Tests for abnormal toxicity are typically carried out using the final bulk or filling lot materials and not unprocessed bulk materials. As some Japanese manufacturers have found that some inactivated whole virion unprocessed bulk samples inoculated into guinea pigs resulted in significant weight loss, it has been recommended that bulk materials be tested as part of the development process. ¹⁵

The JMR for influenza vaccine (whole virion) has been including a leukopenic toxicity test and mouse weight gain test, neither of which is now used in vaccine testing. However, these control parameters are important readouts for vaccine safety and should be included in the testing and design process. ¹⁵

Long‐term stability tests were carried out for both bulk materials and filling lots, including photostability tests for the filling lots. We have confirmed long‐term stability of the bulk and filling lot preparations to be 26 and 15 months respectively. ¹⁷

Electron microscopic analysis and fractionation tests were also used to confirm vaccine formulation stability and virus particle integrity.

The protein content of filling lot supernatants was also measured to confirm antigen adsorption into the aluminum hydroxide gel during the stability testing period, and mouse immunogenicity tests were also carried out to confirm the integrity of the vaccine formulation over time.

Single dose toxicity (rats and dogs), local irritation toxicity (rabbits), repeated dose toxicity (4‐week intervals, rats), reproductive and developmental toxicity (seg II and seg III, rats) and pharmacologic safety (rats and dog) tests were carried out as pre‐clinical tests for the candidate vaccines according to PMDA recommendations.

In the local irritation and toxicity testing, the candidate‐adsorbed vaccine formulation showed similar effects as the commercial‐adsorbed DTaP vaccine suggesting that the test vaccine did not elicit any adverse cutaneous responses.

In the results of the phase I clinical trial, all formulations were well tolerated, no severe adverse events were found and local reactions were much greater in the SC group than in the IM group. However, no differences were found in systemic reactions (Table 2), and the vaccine showed good neutralizing antibody responses between 5 and 15 μg HA/dose. There were no significant differences in neutralizing antibody responses between the SC and IM routes. ⁹

In HA inhibition antibody responses, after the second vaccination, the 15‐μg IM vaccine met two of three CHMP criteria: significant increase in titers (>40%) and mean geometric increase (>2·5 times). However, the 15 μg SC and 5 μg IM vaccines met one of three CHMP criteria (Table 3).

From these results, we have selected the 5‐ and 15‐μg vaccines as vaccine doses for phase II/III trial using SC delivery.

All vaccine formulations were well tolerated with no serious adverse events reported between days 0 and 42. Most of the reported local and systemic reactions were graded as mild and transient. Fewer injection‐site and systemic reactions arose after the second vaccination than after the first in both groups. The highest incidence of injection‐site reactions in each vaccine group were redness, pain, itching, swelling and injection‐site warmth. The incidence of induration was not so high (less than about 10%). The systemic reactions in each vaccine group were weariness, headache and others (e.g. fever: <3% in both vaccine group; data not shown). ⁹

In HA inhibition antibody responses, after one and two vaccinations, the 15‐μg vaccine group met two of three CHMP criteria: significant increase in titers (>40%) and mean geometric increase (>2·5 times), and after two vaccinations, the 5‐μg vaccine group met one CHMP criteria: significant increase in titers (>40%) (data not shown).

Thus the safety and immunogenicity of a two‐dose regimen of adjuvanted 15 μg inactivated whole‐virion H5N1 vaccine given by the SC route in healthy adults were confirmed by this trial.

The vaccine was approved as a mock‐up vaccine and pre‐pandemic vaccine by the MHLW. Both immunization routes, IM and SC were approved.