Dame Deirdre’s still-unanswered questions on swine flu

Dame Deirdre Hine’s report on the 2009 Influenza Pandemic  makes a strong appeal for publication of the analyses on which the Scientific Advisory Group for Emergencies (SAGE) relied, and mentions ‘statistics’ not at all . . . except to welcome (p102) the establishment of a working group to review data collection procedures for the benefit of any future influenza pandemic.
There is praise for safe H1N1 vaccine licensing and production, for the logistics of antiviral delivery, for the National Pandemic Flu Service (NPFS) which operated in England, for doubling of ICU provision ahead of the second wave of H1N1 infections, and for huge efforts by many practitioners both locally and nationally. And thankfulness that H1N1 was, for most, a mild disease.
The National Framework’s planning assumptions in the UK had anticipated a severe pandemic of between 55,000 and 750,000 fatalities, but fewer than 500 H1N1-related deaths occurred. Three key questions go unanswered. 

  1. Why was England’s H1N1 death-rate per million of population about half that of the devolved administrations – did NPFS play a role?
  2. How many lives were saved, and hospitalisations avoided, in UK’s second wave of swine flu by H1N1-vaccination of at-risk groups from late October 2009 – for example, on a population-comparison basis with three winter months of H1N1 in Australia/New Zealand when there was no pandemic vaccine?
  3. Despite immunity and vaccination, how much larger was the second wave of H1N1 in England than its first? In Scotland, there were 172 hospitalisations for confirmed H1N1 to the end of September 2009 (end of first wave), and a further 1,368 to the end of February 2010 (near-end of second wave), which gives a 1: 8 ratio.

The report provides a succinct H1N1-events&decisions timeline from April 2009 to the end of March 2010. The timeline is framed against clinical case estimates (in ratio 1: 2?) - for England, I suspect - for the first and second wave of H1N1 during June to December 2009, a subtle visual reminder that critical ministerial and other decisions in April to May 2009 were poorly, if at all, informed by data (p48).
The plot’s clinical case estimates are not accompanied, as they should be (because different series tell different tales), by incident hospitalisations for confirmed H1N1, incident ICU admissions, and H1N1-related deaths. For example, how many of England’s 342 H1N1-related deaths occurred in the first wave?
In that critically uncertain period, Dame Deirdre notes that, from the end of April to early July, it was Scotland’s Deputy First Minister, supported by her Chief Medical Officer, Dr Harry Burns, who was, daily, the "single authoritative voice" (p134) in Scotland. In England, weekly media briefings were led by its Chief Medical Officer, Sir Liam Donaldson, between 2 July 2009 (start of treatment phase) and 14 January 2010.
Dame Deirdre’s report undoubtedly wins hearts by praising. Willing minds will then take up her recommendations, some trenchant. Among the latter: the underlying science and terminology of “reasonable worst case” should be reworked (R11 & R25); need for adherence (8.37 & 8.38) to existing Cabinet Office guidance on communicating risk that requires transparency in the way assessments are made and decisions taken, and hence her recommended publication of more scientific papers underpinning policy debates and, in particular, that there should have been regular publication of best estimates of spread and fatalities (p9).
Any future SAGE should adhere as closely as possible to the established principles of scientific advice to government and should release its “descriptive papers and forecasts (as distinct from any policy advice) at regular intervals” (R14); and the Government’s Chief Scientific Adviser should provide expert technical briefings to respected scientists not directly involved with SAGE (R15) – as this would enable a wider group of experts to comment in an informed manner on the government’s approach.
That recommendation is particularly important as UK’s independent Scientific Pandemic Influenza (SPI) Advisory Committee was, in effect, stood down for the duration of the H1N1 pandemic by the constitution of SAGE, although some SPI-subgroups worked to SAGE. In particular, SPI-Modelling (p64) was particularly engaged in revising planning assumptions
Dame Deirdre muses: was lower-than-hoped-for uptake of H1N1 vaccine a communications failure to dispel public and professional concerns that the H1N1 vaccine was a) not safe, b) had been rushed into production without the usual rigorous testing; or a pragmatic risk-benefit judgement by the public? She therefore recommends (R22) that the Joint Committee on Vaccination and Immunisation should be asked for advice on vaccination strategies across a range of scenarios, including severe and less severe pandemic viruses, and that this advice should incorporate economic analysis and the views of behavioural scientists.
In reviewing value for money (p22), Dame Deirdre acknowledges that the a priori economic case used a robust methodology applied to prior-to- pandemic assumptions (what these were exactly is unstated, but inevitably will have related to assumed clinical attack rate, together with hospitalisation and death rates among clinical cases); and valued a life saved at £1.6 millions.
How expenditure of £1,200 millions (p155 – excluding research costs) stacks up against fewer than 500 H1N1-related deaths but an un-counted number of lives saved will make for intriguing posthoc economic analysis. Just two approaches to counting/guesstimating the numbers of lives saved are: a) a population-based comparison with the 3 winter months of H1N1 in Australia/New Zealand prior to pandemic vaccine (which might suggest a saving of 350-500 lives in UK’s second wave depending on immune proportion after UK’s first wave); and b) by comparison with seasonal flu in UK (which might mean shortfall of 1,000-2,000 deaths). 
Why the lives saved, for example by H1N1 vaccination, are un-counted was not tackled by Dame Deirdre. However, among confirmed H1N1 cases who were hospitalised or admitted to intensive care units in November 2009, a relative deficit of patients in the vaccine-priority groups – compared to pre-vaccination during UK’s second wave – should also have been looked for. 
Calculation of clinical attack rates was allegedly more complex (4.34) because of large numbers of asymptomatic H1N1 infections which gave subsequent protective immunity (and so reduced second wave susceptibles) and because of pre-existing cross-immunity from H1N1 variants in older individuals. However, the estimated ratio for H1N1 of asymptomatic: symptomatic was not reported; and would have to be much greater than the conventional 1:1 for this explanation to hold sway.
Recommendation 8 is to Chief Medical Officers: that they should jointly commission further work to support key decision-making early in a pandemic. SAGE’s emphasis on modelling was deemed to have reduced the opportunity for a full contribution by other disciplines – notably, those involved in operational epidemiology and clinical care who were better able to understand the virulence of the epidemic (p69).
There was thus a lack of public health challenge – let alone statistical challenge until the Royal Statistical Society took matters up (not referred to by Dame Deirdre) – to the numbers being provided by modellers. Real-time modelling, for which too much had been claimed in advance, became extremely accurate by October 2009, we are told - but we are still denied the opportunity to see for ourselves. But October 2009 was, of course, long past the time when real decisions had had to be made.
(conflict of interest declared:  Professor Bird was nominated by the Royal Statistical Society as statistician-member of UK Scientific Pandemic Influenza Advisory Committee, which did not meet during H1N1)