Sir

In his Commentary 'The contents of the syringe' (Nature 454, 160–161; 2008), Steven Salzberg is highly critical of the World Health Organization (WHO) process for selecting influenza strains to include in vaccines for the coming year. He suggests that predictions could be improved by using sophisticated informatic modelling techniques to interrogate the available sequence and antigenicity data. However, the WHO's expert group responsible for strain selection is now deploying these routinely and its track record has generally been good.

Salzberg argues that the expert group's recommendations should then be opened up for external critique. This is impractical. Strain selection is carried out under huge time pressure, with manufacturers having just six months to deliver tens of millions of vaccine doses. Delays can have very serious economic and political consequences for vaccine producers and health authorities. Decisions are left until as late as possible, usually mid-February for Northern Hemisphere countries, in order to have the widest data set to inform the decision.

Unfortunately, epidemics are sometimes only just beginning at that time. As providers of reagents to standardize vaccine potency, we are acutely aware of pressures inherent in the system, with days making a difference to the delivery of vaccine on schedule. To allow time for critical input, the strain-selection process would need to be brought forward, which would defeat the object by reducing the amount of hard data available to inform the decision.

Salzberg believes that using cell culture, rather than eggs, for production could speed up vaccine production, thus allowing more time for strain selection. The effect, if any, would be small because the time to delivery of final vaccine lots depends primarily on a whole series of quality-control, formulation, filling and packaging steps that are essentially the same whatever the production system.

We must still, therefore, rely on an imperfect process in which a group of experts makes the choice quickly, and as best it can. No doubt there is scope for further improvement, but an open debate before each decision is made would be counter-productive.