Since the outbreak began, researchers have rushed to publish research about the new coronavirus spreading swiftly through the world. On Tuesday, for the second time in recent days, a group of scientists has questioned the data used in studies in two prominent medical journals.
A group of scientists who raised questions last week about a study in The Lancet about the use of antimalarial drugs in coronavirus patients have now objected to another paper about blood pressure medicines in the New England Journal of Medicine, which was published by some of the same authors and relied on the same data registry.
Moments after their open letter was posted online Tuesday morning, the editors of the N.E.J.M. posted an â€œexpression of concernâ€ about the paper, and said they had asked the paperâ€™s authors to provide evidence that the data are reliable.
The Lancet followed later in the day with a statement about its own concerns regarding the malarial drugs paper, saying that the editors have commissioned an independent audit of the data.
Both of the studies relied on an analysis of patient outcomes from a private database run by a company called Surgisphere, which says it has granular information about nearly 100,000 Covid-19 patients from 1,200 hospitals and other health facilities on six continents. Many health care data experts say they knew nothing about its existence until recently.
Both papers were published in May within a few weeks of each other in highly respected medical journals that subject studies to peer review before publication. Both had considerable impact, halting clinical trials of malaria drugs around the world and providing reassurance about the risks of blood pressure medications taken by millions of patients.
But scientists have not seen the large data set that Surgisphere says it has built, and questions about its provenance are rising in scientific circles.
In the open letter to the authors of the N.E.J.M. paper and to the journalâ€™s editor, Dr. Eric J. Rubin, more than 100 clinicians, researchers and statisticians demanded more detailed information about the patient data that served as the basis of the study, and called for independent validation of the work by a third party.
The study was said to analyze 8,910 Covid-19 patients hospitalized through mid-March at 169 medical centers in Asia, Europe and North America. The authors concluded that cardiovascular disease increased their risk of dying.
But the paper also appeared to put to rest any concerns that people with high blood pressure might have about taking drugs called ACE inhibitors: Some people had wondered whether the drugs were playing a role in exacerbating the illness.
Instead, the patients taking these drugs were more likely to survive than those who were not, the authors said. (Other studies have also reported that blood pressure drugs do not make people more susceptible to infection with the coronavirus, and do not increase the risk of more severe illness.)
In the paper published in The Lancet, the authors said they had analyzed data gathered from 671 hospitals on six continents that shared granular medical information about nearly 15,000 patients who had received the antimalarial drugs and 81,000 who had not, while shielding their identities.
The papers concluded that use of chloroquine and hydroxychloroquine may have increased the risk of death in these patients.
The first author on both of the papers is Dr. Mandeep R. Mehra, a cardiovascular specialist and professor at Harvard Medical School. The second author is Dr. Sapan S. Desai, the owner and founder of Surgisphere.
On Tuesday morning, Dr. Desai, who has vigorously defended both the studies and his database, said he and his co-authors on The Lancet study have agreed to a voluntary third-party audit done in collaboration with the journal.
He also said he was arranging the terms of a nondisclosure agreement that would allow the editors of the N.E.J.M. to see the data they had requested.
Dr. Desai had previously said that his contractual agreements with hospitals prevented him from disclosing any hospital-level patient data, even though it was anonymized. â€œSurgisphere stands behind the integrity of our studies and our scientific researchers, clinical partners and data analysts,â€ he said in a statement.
In their letter to the N.E.J.M., critics of the work wrote: â€œSerious, and as yet unanswered, concerns have been raised about the integrity and provenance of these data.â€
The letter points out â€œmajor inconsistenciesâ€ between the number of coronavirus cases recorded in some countries during the study period and the number of patient outcomes reported by the researchers over the same period.
In particular, they said, it is â€œdifficult to reconcileâ€ the Surgisphere data from the United Kingdom with government reports. The paper reported on 706 patients hospitalized with confirmed Covid-19 in just seven of the U.K.â€™s 1,257 National Health Service hospitals.
Yet a high proportion of coronavirus patients hospitalized in the U.K. early on were in London, and no London borough or hospital had more than 100 confirmed cases by March 16, the critics said.
The studyâ€™s numbers on cases in Turkey â€œcannot be correct,â€ according to the letter. The paper reported data on about 346 patients with confirmed cases in three Turkish hospitals by March 15.
But Istanbul University Hospital, one of the largest hospitals in the city, admitted its first Covid-19 patient on March 16, the writers said.
â€œThe majority of patient data in Turkish hospitals are manually entered on paper, and Turkey does not have an electronic nationwide digital database other than for blood tests and prescriptions,â€ the critics added.
â€œMoreover, it is highly unlikely that such clinical data would have been shared with a U.S. company without acknowledgment.â€
Many of the scientists who first raised concerns about the database are involved in clinical trials of chloroquine and hydroxychloroquine, and they were forced to pause the studies for safety reviews after The Lancet study was published.
James Watson, a senior scientist with MORU Tropical Health Network, said his unit had to immediately suspend work on a large randomized clinical trial to see if chloroquine or hydroxychloroquine can protect health care workers exposed on the job to the coronavirus from infection.
â€œI saw very quickly this paper didnâ€™t hold up to much scrutiny at all,â€ he said. â€œWe started wondering, â€˜Whoâ€™s been collecting this data, and where did it come from?â€™ We were quite surprised to see a global study with only four authors listed and no acknowledgment of anyone else.â€
The scientists then turned their attention to the paper about cardiovascular disease and blood pressure drugs that had been published in the N.E.J.M. on May 1. â€œWe immediately thought, â€˜If thereâ€™s something wrong with the database, itâ€™s going to affect both publications,â€™â€ he said.
David Glidden, a professor of biostatistics at University of California, San Francisco, who reads all new publications about Covid-19 antiviral therapies as a member of a National Institutes of Health clinical guidelines panel, said he was immediately struck by the vagueness of the descriptions in both papers.
There is a frenzy to publish research, he added: â€œMedical journals often feel pressure to be relevant and to be carrying the story thatâ€™s going to be talked about, and I think they need to be responsive to the urgency of this pandemic but also to maintain their standards, which require caution.â€