Clearing the air over H1N1 vaccine safety

We have been told, in surveys of dubious validity, that a third of nurses and up to half of doctors would decline to take a vaccine against swine flu. Regardless of whether these figures are right, they do serve to highlight a salient issue: how are we going to design vaccine studies persuasive enough to resolve these doubts?

Nursing Times’ online survey has already been reported here. It produced a 1 per cent response (1,500 respondents from a 140,000 user-base), but the headline figure - one third (of respondent) nurses would decline the pandemic vaccine – certainly won headlines.
 
This week it’s the turn of Pulse magazine. An email survey sent to “several hundred” GPs got a response from 115 of them, of whom 56 said they did not intend to be vaccinated, 48 said they did, and 11 were undecided. Richard Hoey, Editor of Pulse, admits on his website: “I would agree that the survey was relatively small, but it was the third such survey in the last week to find that healthcare professionals are likely to refuse swine flu vaccination in fairly large numbers. A peer-reviewed study in Hong Kong has found the same thing.”
 
Whatever the merits of the surveys, the issue is now in the public domain, and the headlines people have read will influence how they behave. So Professor David Salisbury, the Department of Health’s director of immunisation, needs to ensure that there is transparency about the experimental design of pandemic vaccine studies and the numbers of adults and children randomized (by age-group) per study.
 
This information needs to be in the public domain whether the studies are being conducted by the pharmaceutical industry for licensing purposes (GSK and Baxter make pandemic vaccines) or by independent scientists who have received research funding, such as from UK’s National Institute for Health Research (here).
 
Similar information needs to be made available on earlier studies such as of the AS03 H5N1 flu vaccine which has been assessed in clinical trials of (how many?) adults and (how many?) children. Safety cannot be absolutely guaranteed. Our guarantee is limited by the numbers of individuals who have been exposed to the vaccine in the randomized controlled trials that are the basis for licensing; and in independent trials.
 
In early August, the Chief Medical Office set the scene well for journalists. He explained that fish-oil-based (squalene) adjuvants have been used in flu vaccines for over a decade to stimulate the immune system to produce antibodies more effectively. The adjuvant in the forthcoming GSK pandemic vaccine contains a squalene adjuvant, AS03. Sir Liam pointed out that an AS03 H5N1 flu vaccine had already been assessed in clinical trials in both adults and children – but in how many?
 
Randomized controlled trials in adults and children are designed to determine antibody response (by age) and to rule out serious adverse events of a frequency greater than 1 in X-thousands. Licensing of vaccines can, and must, proceed on the basis of adequate immune response and ‘operational’ safety - in the sense of ‘absence of serious adverse events of types A, B, C in X-thousand persons exposed to the pandemic vaccine’.
 
It is, for this reason, essential that scientists, healthcare workers and citizens alike can readily discover from pharmaceutical companies (GSK and Baxter in particular), licensing authorities (European Medicines Evaluation Authority or Food and Drugs Administration), health departments and research funders (including UK’s National Institute for Health Research) how many people (by age-group) have been, or will be, randomized in vaccine studies; and, according to the study’s design, how many participants are intended to receive two doses of the pandemic vaccine. The total, across all relevant studies, of such participants sets the limit on what is knowable about ‘operational’ safety before a pandemic vaccine is deployed for the protection of priority groups.
 
Pending ethical approvals, it is understandable that study protocols are not yet available for the research recently funded by NIHR. However, it is regrettable that most of the lay abstracts, whose purpose after all is to enhance public understanding of science, fail to document such fundamental elements of study design as the number of people to be randomized and the plausible effect-size that a study has been designed to detect.
 
Vivienne Parry, a lay member of the Joint Committee on Vaccination and Immunisation (JCVI), and Professor Salisbury remind healthcare workers that, by being immunised, they help to protect ‘vulnerable patients’ from the virus and refer to their ‘duty’ to have the vaccine.
 
I respectfully remind JCVI of its ‘duty’ to lay on the line quantitatively the evidence-base we currently have, and that we expect to have, before UK begins its vaccination of priority groups. Let’s not undermine UK science and the challenge that the pharmaceutical industry is facing by being shy about numbers. Large numbers (by age-group) matter fundamentally in underpinning confidence in vaccine safety.
 
Inevitably, our guarantees on ‘operational’ safety will be limited when priority groups for pandemic vaccine are first offered their two doses (21-days apart). What design solutions are there?
 
First, Professor Scott Zeger who delivered this year’s Bradford Hill Lecture suggested that immediate post-vaccine safety could be studied by delivering three doses at 21-day intervals, one of them a placebo (P). Randomization between V V P and P V V allows the safety of the first and third doses (V or P) to be assessed and also ensures that the interval between active doses remains 21-days (which V P V does not).
 
Secondly, funding of general practitioners to administer pandemic vaccine could be tied to bespoke follow-up of a random sample of those whom they vaccinate, to answer questions such as: who contracts H1N1 within D days of initial dose; who develops serious adverse event within S days of initial dose; who develops adequate antibody response within A days of initial dose. Reliance on the yellow-card system for the reporting of side-effects is unlikely to be sufficient.
 
Thirdly, as not all in the priority groups can be immunised simultaneously but all are to be offered vaccination, a wedge design could be used to randomize the order in which healthcare workers – either individually or by workplace – are assigned to a specific week (first week, second week, third week, fourth week etc) for receipt of their first offered dose of pandemic vaccine. This was the design used for the introduction of hepatitis B immunisation in the Gambia more than 20 years ago.
 
Fourthly, H1N1 is currently a mild, moderately transmissible disease. The reproductive number of pandemic influenzas is only of the order of 1.8, which means that to bring the reproductive number below 1, and hence to control the epidemic by immunisation alone, we should need to immunize effectively at least 0.8/1.8 = 44 per cent of the at-risk population. If, somewhat pessimistically, adequate immune response were achieved in only 80 per cent of vaccinees, then we’d require uptake by at least 1.0/1.8 = 56 per cent of the at-risk population to control H1N1’s reproductive number. Thus, the monitoring of uptake rates matters.
 
In summary, UK has a “duty” to healthcare workers and others to design well, and report publicly, its roll-out of pandemic vaccination to priority groups. Priority groups include healthcare workers who understand very well both the virtues of randomisation and the need for UK to add informatively to the international evidence-base (by age-group) on pandemic vaccine safety. I suggest that healthcare workers need no lessons in “duty” to their patients. Rather, they are entitled to the best information scientific method can provide.