Thread by @peripatetical

A large group of us have expressed concern about The Lancet HCQ/CQ study (10.1016/S0140-6736(20)31180-6). @TheLancet
zenodo.org/record/3862789…

There are many criticisms and anomalies, but a few notes about the Australian data. The authors reported 609 admissions and 73 deaths in 5 Australian hospitals on 21 April.

Curiously, no Australian data were included in a previous paper on cardiovascular disease by the same authors in the @NEJM (10.1056/NEJMoa2007621) using the same database to 28 March

In Australia, 72 people were reported to have died from COVID-19 on 21 April.
covid19data.com.au

On 19 April, the Australian government reported that 810 people had been admitted to hospital but 167 were still in hospital.
www1.health.gov.au/internet/main/…
infogram.com/1p0kn5222knxnk…

The authors responded by saying that one hospital had been misclassified as Australian, and there were 68 deaths.
theguardian.com/science/2020/m…

However, the state breakdown of deaths was NSW 29, VIC 15, TAS 8, WA 7, QLD 6, SA 4, ACT 3. So even if the four hospitals had treated for all the people who had died in the states with the most deaths (unlikely), this wouldn't be possible. covid19data.com.au/deaths

Additionally, the authors reported that 49 patients had received CQ and 50 had received HCQ. CQ isn't currently available in Australia.

CQ can be imported under the Special Access Scheme in Australia, but it would seem unlikely that large numbers of patients would have received CQ when HCQ is available.
tga.gov.au/form/special-a…

These anomalies cast doubt on the veracity of the database - it may well be a simple error that explains all this (Austria!?), but an independent review and audit is required.
news.com.au/entertainment/…

There have been other observational studies of HCQ and CQ. But these are all likely to have various biases - treatment tends to be given to those more seriously unwell, but patients need to survive long enough to receive it.

Ultimately, we'll need a randomized controlled trial to know if CQ or HCQ have any effect on the time to recovery or mortality. It would be a pity if these trials were stopped in response to these data.

Update: The Lancet have published an erratum. The authors now report 63 patients and 3 deaths from Australia, which is plausible.
thelancet.com/lancet/article…

The one hospital that misreported being in Australia therefore must have had 546 patients. There are enough inconsistencies and issues that suggest that an independent audit and review is still required.

Zenodo seems to be down at the moment, here's another copy:
documentcloud.org/documents/6933…