On 2020-04-24 09:57:00, user Philip Davies wrote:
Well, well well,
This pre-print would make a good script for an episode of Columbo.
The retrospective analysis, as presented, leads the reader to just one conclusion in a bazaar of many possible conclusions.
I am even starting to have sympathy with D. Raoult and his team. I note his hot tempered response to this paper, where he lists two enormous factors that should be considered when wrestling with the data: the fact that the HCQ and HCQ & AZ cohorts were a sicker crowd (he lists lymphopenia) and that the sickest of the non-HCQ ventilated patients were then given HCQ (plus AZ in most cases) in a desperate last bid only for most to die.
Raoult's point is certainly valid.
We must remember that for most of the study period the use of HCQ was "ex-license" on a compassionate basis only. This means only the sickest patients got it. Remember also that this is a retrospective analysis, therefore observational. It was not run as a therapeutic trial. On the other hand, the use of AZ was already accepted (hence 30% of the non-HCQ cohort got it anyway).... although do be aware that by this time there had been quite a lot of focus on potentially dangerous QT lengthening when HCQ and AZ were used together in very sick patients.
The HCQ cohort was, across all key determinants, the weakest and sickest group (it had the poorest prospects looking at age, ethnicity, smoking status, congestive heart failure, peripheral vascular disease, cerebrovascular disease (strokes),dementia, COPD, Diabetes (with and without complications)! ... and indeed, the HCQ and HCQ & AZ cohorts did have 100% more lymphopenia than the non-HCQ group.
BUT, the big asymmetric issues become obvious when we look at the pre- and post- ventilator numbers.
In terms of patients discharged without needing ventilation, the "victorious" non-HCQ group performs poorer than the 2 treated groups. This despite having a better prognostic baseline. But the results for this group change dramatically (for the better) when we look at the outcomes of ventilation. 25 ventilated patients came from this group.... but 19 of these 25 patients were then started on HCQ or HCQ & AZ after ventilation was started. It is screamingly obvious that these would be the sickest patients in that group: they were given such compassionate drugs in extremis. So having ejected 19 of 25 ventilated patients into the other cohorts, the non-HCQ group only had 3 deaths from its remaining 6 ventilated patients.
The numbers of ventilated patients in the other cohorts (HCQ and HCQ & AZ) were thus substantially inflated with these new super-sick patients, who mostly died.
There really can be no conclusion at all when looking at a study of this nature without knowing much more about individual clinical conditions and guiding principles behind clinician's decision making. It's still possible to make some reasonable assumptions:
If I were Columbo?... I would say the non-HCQ cohort contained patients of extremes, with the best and worst potential. The worst would have been the very frail (malignancy and or congestive heart failure maybe ... see the stats), who probably were earmarked for 'supplemental oxygen' only from the very start. Such patients would not have been suitable for compassionate use of non proven drugs (remember, most of this came before the "emergency use" edict by FDA). This would explain the number of non-ventilated patients who died in this group (they may have been given AZ only, not being a controversial drug, but otherwise they did not get any significant interventional therapy). These patients would have had significant chronic disease and very poor obs/indices (including lymphopenia). But given that this cohort had, overall, a better starting prognosis than the other two groups, it means that the remaining patients in the group were promising candidates for survival (with better obs/indices). Such patients, not being part of a clinical trial, would not have been offered HCQ on a compassionate basis unless they got dramatically worse .... and of course, the ones who did get worse on the ventilator were started on HCQ (& often AZ as well) and thus swapped into the HCQ / HCQ & AZ cohorts.
If we can understand that, then we might start to think that in fact HCQ & AZ is the best performing cohort with the other 2 vaguely distant. But this is being unfair to the HCQ cohort:
The reason that a sick patient would be given one experimental drug on a compassionate basis (HCQ) but not have a rather less experimental drug further added (AZ), can really only be explained by considering risk versus benefit. A clinician would choose to use HCQ because the patient was particularly sick. The clinician would only add AZ if it was felt that this was worth the risk.... but a particularly sick patient with significant cardiovascular disease (the HCQ contained the most CVD risk) might then die of a more abrupt arrhythmia through adding yet another QT lengthening drug. I dare say the clinicians were tempted to make some "Hail Mary" plays, but we must remember, these patients were not part of an ongoing trial, these drugs were "ex-license" for compassionate use only and clinicians were still accountable for responsible actions. So for those particularly sick frail patients, it wasn't worth the risk.
I am pretty sure that the HCQ cohort (which had pretty good pre-ventilator stats) crashed badly because it was loaded with the sickest patients .... patients that were too sick to risk adding AZ.
So, the findings of this retrospective analysis are, in my opinion, likely to be incorrect.
I believe I can confidently state that:
- The HCQ cohort started with the sickest patients and had even more of the sickest added during ventilation. Some were too sick to risk the addition of AZ to existing HCQ.
- The HCQ/AZ cohort also had some very sick patients (again with more additions during ventilation).
- The Non-HCQ cohort had the best prognosis overall from the very start (although likely a polarized mixture of the most frail and the most promising)... and then its stats got even better when it jettisoned its sickest ventilated patients into the other 2 cohorts.
It is almost impossible to reach a conclusion from all this. BUT, the most likely finding is NOT that adding HCQ delivers a worse outcome than standard treatment. In fact, if we look at the pre-ventilator stats, the addition of HCQ might actually have provided considerable benefit to a particularly sick group of patients. Whether or not the addition of AZ to HCQ adds benefit is also unclear ... although my 'swingometer' is pointing slightly more to benefit than harm.
Once again. I suggest that a robust study into prophylaxis and early treatment (using sensible safer doses adjusted for pulmonary sequestration) will deliver the most interesting results for CQ/HCQ.
Dr Phil Davies<br />
Aldershot Centre For Health<br />
http://thevirus.uk
EditView in discussion<br />
Discussion on medrxiv 3 comments<br />
medrxiv viewer<br />
Philip Davies<br />
Philip Davies 4 days ago<br />
The low dose arm of this study is worth following.
The big problem for this study is comparison. It really has not defined the control population at all. The Italian and Chinese references are entirely different. Even the 2 Chinese populations referenced had massively different outcomes because the populations examined were different.
The Italian mortality rate was actually similar to the overall study average here (but much higher than the low dose arm). The Chinese study involved all patients admitted to the two hospitals ... that included a majority of patients with moderate ("ordinary" as the Chinese class it) disease severity. The patients in this Brazilian study were regarded as severe or critical ... such patients (looking at worldwide stats) would attract a mortality of 30-40% plus.
This is the most important factor. Do not compare apples with pears. So far this study points the "swingometer" in favor of benefit versus harm for the use of HQN in patients with advanced disease.
Once again however, we are looking at the potential impact of an orally administered drug to patients with advanced disease. That's a big ask.
For CQ and HCQ the most interesting results will likely come from studies looking at prophylaxis and early treatment (using safe doses, not silly high doses with added drugs that also lengthen QT). We can't yet guess how they will pan out.
Dr Philip Davies<br />
GP<br />
Aldershot Centre For Health, UK<br />
http://thevirus.uk