Obfuscation and evasion: Astra shows how not to disclose data

It is a good idea, in the middle of a global pandemic, to be fully transparent. Yet Astrazeneca’s disclosures around its Covid-19 vaccine AZD1222 have tended to obfuscate and evade, even before the US NIAID's extraordinary claim that yesterday’s pivotal US data might have relied on outdated information.

Though the NIAID does not spell it out, the suggestion is that the result, touted as highly positive, might not look as good had it included more up-to-date findings. The concerns could of course be groundless. But confidence in AZD1222 has been seriously undermined, and if vaccine hesitancy now rises among the public then the global Covid-19 recovery effort will be worse off.

Doubts have swirled around AZD1222 for some time, and perhaps this is why Astra decided to go in with guns blazing yesterday. But the group must surely have been aware of the damaging effect of being unclear, even if it did not imagine that the wheels would fall off within 12 hours.

Crisis

Yet, with a clear disagreement between the NIAID and Astra now playing out in the full public glare, this is the crisis the group’s PR team is having to firefight today.

In a statement this morning Astra said preliminary assessment of the full primary analysis of AZD1222's pivotal US trial was “consistent” with yesterday’s interim numbers, taken at a February 17 data cut-off. “We intend to issue results of the primary analysis within 48 hours,” the group stated.

This was a brief response to the NIAID’s possibly unprecedented intervention hours earlier. Overnight the US body claimed that it had been informed by the trial’s data-monitoring board late yesterday that possible inclusion of “outdated information ... may have provided an incomplete view of the efficacy data”.

On the face of it the data were highly positive: 79% efficacy in preventing symptomatic Covid-19, 100% protection against severe disease and hospitalisations, and no increased risk of thrombosis. But on a press call the group admitted that more Covid-19 cases had accrued since this interim analysis.

This is not unusual, and indeed the trial’s interim stopping criteria specified around 75 cases, while 141 were actually detailed. So the question is how many cases above the 141 symptomatic – and five severe – there actually were, and what effect including these would have had.

No comment

Astra’s PR strategy yesterday did not help: the company was asked several times to break down the cases by active and placebo cohorts, and refused to do so. It only grudgingly accepted that there were five severe cases, and that all occurred on placebo.

These are by no means Astra's only failings in developing a vaccine wounded by links, later dispelled, to thrombosis that had led several EU countries to suspend it.

Its pivotal US trial, for instance, had spent weeks on hold, and the reason – a case of transverse myelitis – was disclosed not in a public forum but at a private investor call. It later took the Astra/Oxford team a long time to accept that the transverse myelitis was “possibly related to vaccination”.

And it remains a mystery why Brazil/UK data with a half-strength AZD1222 priming dose gave 90% efficacy whereas a full-strength one saw just 62%.

Indeed, it was this finding that highlighted major flaws in the conduct of that first big AZD1222 trial: the half-strength dosing was actually a mistake amounting to a protocol violation. Again, such details were allowed to dribble out piecemeal.

Astra now desperately needs to take control. Even the NIAID accepts that AZD1222 is very likely a good vaccine, but Astra’s failings might have done it irreparable harm. In 48 hours the truth will out. Maybe.