Halting hydroxychloroquine trials based on faulty Lancet study is 'knee jerk reaction', say CSIR, IGIB, CMI

By Suchitra Mukherjee
New Delhi [India], May 30 : Council of Scientific and Industrial Research (CSIR) Director-General Shekhar Mande, Institute of Genomics and Integrative Biology (IGIB) Delhi Director Anurag Agarwal and Chennai Mathematical Institute (CMI) Chennai Director Rajeev Karandikar have written to the World Health Organisation (WHO) and asked the global health body to resume trial of hydroxychloroquine (HCQ) for the treatment of COVID-19 patients as early as possible, asserting that the Lancet's study is not correct.
In an email to Lancet paper editor Richard Charles Horton as well as WHO, Mande stated the study published in Lancet on HCQ and chloroquine and its effect on COVID-19 is an observational study with statical parameters and analysis, which is not sufficiently good.
It also stated that the decision of WHO to temporarily suspend quality HCQ trials based on findings of the Lancet study appears to be a "knee-jerk reaction".
Mande said that the study design and statistical interference are not right and therefore, it does not deserve any attention of the readers.
"In our considered opinion, the decision of WHO is questionable and the statistical analysis is not correct. Lancet paper is not sufficiently rigorous, rather it is faulty and therefore, the decision of the WHO to temporarily halt the study is not correct. We have written an email to the WHO that the decision is not right and the HCQ clinical trial must resume as early as possible," Mande said.
The email refers to the article in Lancet "Hydroxychloroquine or chloroquine with or without a macrolide for treatment of COVID-19: a multinational registry analysis", which present an analysis of the hospital registry data on the effect of HCQ on COVID-19 patients.
According to the letter, the article published in Lancet appears to give a sense of legitimacy to the findings and consequently has led to the suspension of HCQ trial by the WHO.
"However, the study has several limitations as the authors have themselves acknowledged in the article. Thus, using this study to halt the use of HCQ or chloroquine in the ongoing controlled trial seems questionable," it said.
"The article represents an observational study. Unlike randomised blinded trials where subjects are chosen randomly to receive treatment, to minimise differences between treatment groups in this study, there has been no such prior design. The methodology of the paper mentions that all patients who were hospitalised between December 20, 2019, and April 14, 2020, with PCR confirmed COVID-19 infection were analysed," read the letter.
The research bodies said that first reported symptoms of the diseases was a cluster of pneumonia cases which were recorded on December 31, 2019, in China's Wuhan and the WHO published the first disease outbreak news on the new virus only on January 5, 2020.
"Moreover, it was only on January 20, 2020, that the first case was recorded in the US. Thus, scanning the records of hospitalisation from December 20, 2019, appears to be unjustified. While it is not expected to make any difference since PCR confirmation would not be there for non-COVID cases, it is worth noting that the number of patients and the number of deaths in this study are inconsistent with public health records of the same, often exceeding them despite surveying only a subset of healthcare organisations," read the letter.
They said that this raises concern for errors in data extraction from electronic records.
"Even if the data were correctly extracted, severe flaws remain. The main problematic area here, as in all observational studies, is the baseline differences between groups and application of statistical methods to adjust for them," the letter stated.
"Clearly, in this Electronic Health Record (EHR) data-driven study, without any inputs of local treatment protocols, we do not know whether the patients, who were administered chloroquine/hydroxychloroquine were chosen based on -- the severity of symptoms, the number of symptoms reported/observed or location-specific: some hospitals gave it to all (or most) and some gave to few. If so, then the conclusions have no value, because the confounding factors that this choice introduces cannot be statistically corrected using available data," it said.
The research bodies said that on various parameters, the patients who were given chloroquine/ hydroxychloroquine seem to be doing far worse.
"Table 2 shows that the control group has the lowest baseline severity markers, BMI, proportion of males, coronary artery disease, congestive heart failure, diabetes and hypertension amongst all groups. These represent all the major known risks and while individual differences are small and may lack statistical significance, it appears likely that physicians were selecting for sicker patients while assigning treatment with HCQ/CQ," the letter said.
"This is not unexpected here and highlights the well-known problem of statistical comparisons of unbalanced unrandomised groups in observational studies. Conventional statistics test the null hypothesis, which is automatically rejected if the groups are likely to be dissimilar. Little value is then added by confidence intervals and p-values in determining whether differences are causally related to the intervention label of the group," it said.
"Overall, this suggests that receiving the treatment was based on the severity of symptoms. While this does support the lack of high effectiveness of HCQ/CQ therapy, the decision of WHO to temporarily suspend quality HCQ trials based on findings of this study appears to be a knee jerk reaction," it further said.