鶹ý

Surgisphere Data Used in Two COVID Studies Called Into Question

— Scrutiny prompts Lancet, NEJM to conduct independent data reviews

MedpageToday
The Surgisphere logo

Discrepancies in two COVID-19 studies published in major research journals have caught the attention of scientists around the world, leading to calls for investigations into the company that provided the study data.

Questions have been raised about , a data collection company founded by Sapan Desai, MD, PhD, that says it gathers and stores de-identified electronic health record data from 1,200 healthcare organizations in 45 countries.

Alarms were first sounded about a recent Lancet study on hydroxychloroquine (HCQ) and chloroquine for COVID-19 that appeared to have surprisingly high mortality rates overall. Researchers were also concerned that Surgisphere's number of COVID-19 deaths in Australia exceeded the country's published total.

Similar questions were raised about a that found common blood pressure medications were not associated with in-hospital death among hospitalized patients with COVID-19, assuaging concerns about a possible harmful association. In that study, the number of COVID-19 cases Surgisphere counted in Turkey in mid-March was vastly greater than official numbers at the time -- by a factor of about 80, according to a , of Grant Family Medicine in Columbus, Ohio.

"At this point, there appear two viable options: the analysis and/or data were of such poor quality that it would render the studies unsuitable for publication, or the studies are derived from falsified data," Johansen told 鶹ý in an email. "Two of the world's most prestigious medical journals published studies by the same group of authors that are, at best, of no value with numerous obvious errors."

Scrutiny from the medical community has moved editors from both and to publish "expressions of concern" announcing investigations into the database. The Lancet had also previously regarding the Australian deaths.

"Although an independent audit of the provenance and validity of the data has been commissioned by the authors not affiliated with Surgisphere and is ongoing, with results expected very shortly, we are issuing an Expression of Concern to alert readers to the fact that serious scientific questions have been brought to our attention," wrote the Lancet editors.

Desai, who was a co-author on both papers, did not respond to a request for comment Tuesday, but he previously told 鶹ý the company is reviewing the Lancet analysis to ensure there are no issues with the data. Surgisphere claims to extract data from participating hospitals' electronic records in near-real time. In contrast, Desai said, official reports from public health agencies often lag considerably. This would lead to substantial discrepancies when a disease outbreak is growing rapidly, as was the case during the periods examined in the two studies.

Other experts were skeptical, however.

Adrian Hernandez, MD, cardiologist and director of the Duke Clinical Research Institute at the School of Medicine in Durham, said collecting data from hundreds of hospitals would be nearly impossible during the pandemic as COVID-19 testing, diagnoses, and EHR coding have changed at a rapid pace.

"Extend that across six continents with huge variation in terms of accessibility of electronic health data, that seems also very challenging," Hernandez told 鶹ý.

Currently, Surgisphere lists (though Desai he had 11 employees), many of whom appear to have joined the team in the past few months.

"Right now there is zero proof that some highly-engineered software is capable of this," said James Todaro, MD, of Columbia University, who . "It gives the appearance of a tech startup trying to create the impression it's bigger than it is, as opposed to a database managing highly sensitive patient information for over 1,000 hospitals."

Critics asserted that the amount of work necessary to obtain and operate the complex dataset they used in their analysis is incomprehensible for a company of that size.

"In both the NEJM and Lancet papers, the feasibility of pooling so much data across 1,200 hospitals ... is unfathomable," said Boback Ziaeian, MD, PhD, a health services researcher at the University of California Los Angeles, who studies cardiovascular outcomes.

Scientists immersed in COVID-19 literature often anticipate studies months before they are published in major journals like The Lancet or NEJM. Hernandez said a number of organizations, including large public companies, health tech, government agencies, and academic consortia, have been searching for large-scale COVID-19 data like this.

"This resource popped up out of nowhere, which was a complete shock in terms of how they did this," Hernandez said.

No evidence of data reliability has been released to date, and Hernandez said there is a need for a more granular geographic breakdown of data at the hospital level, or, at minimum, the regional level.

"There shouldn't be any reason why you couldn't do that in a de-identified way," Hernandez said. "Whenever someone raises these kinds of questions, the first step is getting to full transparency. That wasn't attempted at all early on."

The substantial increased mortality and occurrence of cardiac arrhythmias associated with HCQ and chloroquine observed in the Lancet trial was and led the World Health Organization (WHO) to pause recruitment for the HCQ arm in its Solidarity trial, . Other trials in Europe were halted as well, according to an signed by a group of more than 180 medical researchers, statisticians, and ethicists around the world.

"Given the enormous importance and influence of these results, we believe it is imperative that [t]he company Surgisphere provides details on data provenance," they wrote to Desai and co-authors, also asking for independent validation of the analysis and open access to all data sharing agreements Surgisphere claims to have with governments around the world.

Desai created Surgisphere in 2008 when he was a surgical resident at Duke University in North Carolina. In addition to the used for the two studies, he also created QuartzClinical and Apex Testing. QuartzClinical accessible through the "Collaborative," as it's referred to in the Lancet study, in mid-March. Apex Testing is a test preparation company based in Chicago through which Desai .

On its website, Surgisphere cites "" as a major reason that hospitals participate in the registry.

"While our data use agreements with these institutions prevents us from sharing patient level data or customer names, we are able to complete appropriate analyses and share aggregate findings to the wider scientific community," it states.

In response to questions about the data's validity, Lancet paper lead author Mandeep Mehra, MD, of Brigham and Women's Hospital in Boston, said in a statement to 鶹ý that co-authors independent of Surgisphere on both the Lancet and NEJM studies have engaged independent, third-party auditors to review data.

Mehra added that "the data available through Surgisphere were leveraged to provide observational guidance to inform the care of hospitalized Covid-19 patients amidst this global pandemic." Until findings from randomized clinical trials are available, "given the urgency of the situation, using the available dataset was an intermediary step."

  • author['full_name']

    Elizabeth Hlavinka covers clinical news, features, and investigative pieces for 鶹ý. She also produces episodes for the Anamnesis podcast.

  • Amanda D'Ambrosio is a reporter on 鶹ý’s enterprise & investigative team. She covers obstetrics-gynecology and other clinical news, and writes features about the U.S. healthcare system.