Before the hearings, Didier Raoult and Karine Lacombe this week by the parliamentary commission, a “pre-publication” of the APHP on the website MedRxiv of the university of Cornell (USA), it could be just one more episode in the turf war between the clinicians of the Paris and the Pr Raoult. The interpretation of this study is surprising, especially since it comes out at an opportune as the battle of the dual therapy by hydroxychloroquine and azithromycin the Pr Raoult has raged, and continues as we see here, to defray the chronicle. Not to mention the statements of the Pr Perronne in his book does it an error that they have not committed ? Observational medicine, which is based on the best possible care is systematically provided to every patient, practiced by the IHU Marseille has yielded positive results since the number of hospitalizations and subsequent deaths has been greatly reduced compared to the other French departments (see article on the 1st of June). These results and methods have been questioned by a part of the medical community questioning the terms of the medical practices of the IHU, opposing two conceptions of medicine. The empiricism of science, observational accompanied by humanism (each patient is treated the best way possible), against the realism cold of randomized controlled clinical trials, with studies Big Data, which enables any deviations to be fraudulent (e.g. study The Lancet, removed since). The randomized trials are far from being the panacea of scientific rigor as it has been demonstrated with overdose of hydroxychloroquine not actually justified in the test Recovery. Note also the silence of the French study Discovery of which we do not hear absolutely more to talk about, although we had been presented by the government as the miracle cure opposed to the supposed amateurism of the Pr Raoult.
The instrumentalisation of the site MedRxiv for political purposes
Before proceeding to a critical analysis of the “pre-publication” of the article deposited on the fruit of a consortium collaborative research AP-HP/Universities/Inserm on the COVID-19 let’s make a reminder in brief of essential points. First of all, it is dishonest to talk of pre-publication in relation to this text. It is an abuse of language, because the term pre-publication relates only to articles accepted for publication by the system of peer reviewing (i.e. peer review) of which the manuscript is put in public access to the advance until it appears officially in an issue of the scientific journal where it was accepted.
Remember that Cornell university has established a website for the electronic deposit of scientific texts, usually called arXiv.org, which allows you to drop any result in the form of e-print (improperly confounded with pre-print) in absolutely all areas of science. You just need to have an affiliation with a university institute, and in this case, a simple email address is enough.
The initial purpose of this site archive of scientific documents is multiple. It is a program of Cornell university which attempted to bring together all the scientific papers produced in the world irrespective of the fact whether they are published or not. In practice, to users the interest is to be able to preserve the paternity of a discovery by proving its anteriority in relation to any publications of other research groups on the same topic.
Let us recall two examples of spoof resounding discoveries in the past. In 1986, Johannes Bednorz and Karl Müller discover superconductivity -238,15 °C in materials structure, perovskite copper-based lanthane16 (Nobel prize physics 1987). They had, however, failed to be stripped of their discovery by an american competitor, the Pr Chu, who was a reviewer pair from the article they had submitted for publication. Professor Chu took the opportunity to submit the same article in 24 hours after in another journal.
A similar story, even more famous, had arrived at the Pr Montagnier with the photo of the aids virus that had been subtilisée during a working visit to the USA in the laboratory of professor Gallo, the great virologist american of the time. It wasn’t that the Nobel prize eluded him if it is not the honesty of the president of the Academy of Sciences American who later acknowledged the fraud. The dispute had been settled definitively in 1987, during an official visit of Jacques Chirac (prime minister) in the USA where he had met Ronald Reagan.
This being said, we affirm here that the filing of the text of the consortium AP-HP on the site MedRxiv June 16, 2020 has been done for reasons of rivalry and undeclared war against the Pr Raoult, who used a very little expensive, in blatant contradiction to the dogma developed by a lot of clinicians and methodologists in paris. This rivalry is very far from the interests of patients and the search for truth. The deposit of the text on MedRxiv is obviously not the protection of a discovery, but a autopublication by the consortium AP-HP of results without going through the system of peer-reviewing. It is thus similar to a political maneuver. The text is accompanied, as all the documents filed on MedRxiv, a note could not be clearer :
We list at the end of this article a series of open-ended questions, are made to authors on inconsistencies and contradictions that the text raises in the first analysis.
We point out that this article, involving treatment with HCQ plus AZI, is not the only self-published recently on MedRxiv. A study of Big Data, filed on April 8, 2020, conducted on more than 300,000 patients who received the combination therapy HCQ + AZI (since 2000) demonstrated, in a statistically significant way, by meta-analysis that this treatment doubles the risk of heart infarct in relation to the decision only to HCQ ; and increases by 22% the risk of angina. If you look at the figures more closely, this corresponds to 31 patients (HCQ + AZM) against 16 patients (HCQ) on every 10,000 patients treated without learn nothing and on the dosage or the length of the treatments administered.
Moreover, by definition, any effective treatment is also toxic and that the art of medicine is to determine when a treatment may be prescribed in the interest of the patient and under control of the doctor.
A French study of the AP-HP on 4642 patients Covid-19
This pre-publication dated June 16, 2020 seems to be of interest, since it comes shortly before the hearing of Karine Lacombe on Thursday. However, it is important to consider that this is primarily a study of the AP-HP which has opposed wind standing dual therapy Raoult on the grounds that there needed to be a randomised clinical trial to prove its effectiveness while the approach of observational Bouches-du-Rhône enables you to issue a vision far more positive.
The objective of this study is to evaluate the clinical efficacy of hydroxychloroquine orally (HCQ) with or without azithromycin (AZI) for the prevention of death or reduce the length of hospital stay. This retrospective study focuses on 4642 patients Covid-19 French hospital who received hydroxychloroquine, with or without azithromycin. (The study by Lancet was done the same way except that the latter does not mention the precise provenance of medical records)
Interesting conclusions that leaves however place for the many questions
Using a large sample of patients Covid-19 hospitalized in 39 hospitals in France and of the methodological approaches to be robust, we found no evidence of the effectiveness of HCQ or HCQ combined with AZI on mortality at 28 days.
Our results suggest a possible excess risk of mortality associated with HCQ combined with AZI, but not HCQ alone. Rates of return at home significantly more levels have been observed in patients treated with HCQ. A discovery that will no doubt try to validate in studies replicative.
Important precautions to be taken in the analysis and interpretation of results
Data which is essential to estimate the severity of the disease in patients (O2 saturation, and mechanical Ventilation), are missing for more than 2000 patients, nearly half of the cohort. The distribution of this lack of data is not specified, which does not allow to assess the homogeneity of the advanced state of the disease in patients in the different groups. In addition, the group HCQ and the group HCQ+AZI, show the amounts of C-reactive protein (mg/L) significantly higher than those of the control group (77,3 mg/L ; 85,5 mg/L and 65 mg/L, respectively),
Another major element in 39% (88 of 227) of the group HCQ+AZI began to take the treatment in the icu or at the entrance in the icu. A stage of the disease where it has been said many times that this treatment is not useful at all and this skews completely the conclusions by the following
The treated group is more at risk than the control group vis-à-vis the disease. In the group of patients HCQ / AZI compared to the control group, it can be noted that there are more males (69.6 percent vs. 57,1%), more obese (26% vs. 12.3%), more asthma (13.2 per cent vs 7.4 per cent) more patients of the lung (11.9 per cent vs 6.9%), and more of hepatic failure (11% vs. 4.2%).
In statistical terms, and, more particularly, in the multivariate analysis there is a lot to complain about both in the use of the tool of the statistical tests used are tests bivariate so that the hypothesis to be tested is likely to HCQ or HCQ+AZI is superior to ” no treatment “.
Despite these observations of funds, a detailed analysis allows a different vision of the authors ‘ conclusions
On the basis of the results provided, on the mortality of the two treatments HCQ alone and HCQ+AZM provide results that are statistically significant compared to the arm without treatment. There are :
This is a prosaic and statistically significant, in human terms :
Unadjusted clinical outcomes
The only element which remains non-significant, it is the mortality rate on the patients who had to be transferred to an intensive care unit in the first 24 hours. One hypothesis could be related to the fact that the patient’s condition was sufficiently severe or comorbidities may have played a role. For this it is necessary to analyze the ECG and other data at the individual level in order to really be able to make a scientific interpretation.
Primary and secondary outcomes
This interpretation is very important as these results would help to reconcile the two clans “pro dual therapy” versus “pro-RCT” (randomized clinical trials); the results of observational quantified by the IHU Marseille is reflecting finally in a statistical analysis.
The commission of inquiry will need to ask these questions to Karine Lacombe during his hearing and beyond her role as an investigator will be surely to play a role of police officer if necessary at this critical period. Make the pledge that a multivariate statistical analysis taking into account the conflicts of interest will not be necessary for the commission to straighten out the results.
Appendix : open-ended Questions to the authors of the article submitted to the website Medrxiv, fruit of a research collaboration, AP-HP/Universities/Inserm on the COVID-19.
with the title :
“Hydroxychloroquine with gold without azithromycin and in-hospital mortality or discharge in patients hospitalized for COVID-19 infection: a cohort study of 4,642 in-patients in France”
What do you mean by “large non-selected population of inpatients hospitalized for COVID 19 in 39 hospital in France…” ? The sections Data source and Data acquisition raises total opacity on the selection criteria of 4,642 patients that you enter in this observational study. How many patients in total were hospitalized for infection COVDI 19 by the 39 hospitals in your research consortium between the 1 February and the 6 April ?
Why don’t you set up with a computer tool, ensuring a random sampling, an arm of the reference with a sub-group of a few hundred (for example around 500) of patients, among the 3792 arm of reference, co-measured with the arm HCQ and HCQ + AZI and sharing the same demographic characteristics and comorbidities of baseline (baseline characteristics) ? There are visibly to the naked eye a systematic bias of the patients in arm HCQ and even more marked in the arm HCQ + AZI to risk factors (for example among others : male + 12.5 per cent ; obesity is a + 13.7 % ; diabetes + 6.8% to name a few ! On 14 characteristics listed in Table 1 that are well recognised as risk factors for development towards a lethal disease 10 are imbalanced between arms, with treatment and without treatment… do you think that this could guarantee a statistical treatment of reliable your data even by using a powerful tool of statistical analysis (AIPTW), allowing a priori to straighten out some problem of bias estimate of treatment ? Wouldn’t it have been more appropriate to establish a reference arms balanced given that it was possible, then, that the method AIPTW present limits of reliability related to the modeling underlying data.
The approach explained in point 2 (above) is still possible to check the validity of your results obtained by the method AITPW. Are you willing to verify this ? Or let other research teams such as that of professor Raoult verify this by sharing your data ?
Why return values are not statistically significant, as established by the treatment AIPTW, in Table 3, which is used to draw the conclusions of this observational study ?
Should we conclude that the data were not significant (that is to say, those which are not indicated in bold) must not be taken into account ?
That means the fact that the results that cancel, partially, the conclusion to the effectiveness on treatment HCQ and totally on the effectiveness of HCQ +AZI (measured in a univariate, raw data) are not statistically significant ?
How can you explain the method AITPW reversed (cancelled) the favourable outcome for treatment with HCQ plus AZI (Table 3) ? Can this be related to the imbalance between the arms, or to confounding factors ? What are they ?
How is it that the arm HCQ spring when the same asset in relation to the improvement of the healing after treatment AIPTW and not the arm HCQ + AZI?
Why not present the results of the multivariate analysis with the Cox model prior to treatment by the method IAPTW ?
Why use the value of “cutoff” of 1 day (<1 day or > 1 day) in the table 2 ? Do you not think that using a value of 4 days could change the look of the results and their interpretation in the table? What are the values with a cutoff of 4 days ? We believe that the cutoff of 1 day does not allow to put in evidence the effect of the treatments HCQ and HCQ + AZI, because it is not higher than the upper values of the confidence interval for HCQ 1.18 update [0.15;3.49]; and HCQ + AZI 0.37 days [0.07;2.26] ), but it is to the absence of treatment with 0.16 day [0.00;0.80] )…
Is it that the patients in the arm HCQ and HCQ + AZI cover exactly the same period as those of the reference arms. What dates are the most advanced in the time correspond to the last patient hospitalized in the arm HCQ and the last patient hospitalized in the arm HCA + AZI ?
Have you taken into account the fact that the ratio between the patients in the icu and hospitalized patients has continued to decline during the epidemic (see curve determined by reference to the official data below)? This could induce a bias in the results of your study ?
Covid 19 – France resuscitation
Author(s): The Citizens ‘ group for FranceSoir