— Statistician alleges data in study led by Paul Marik, MD, were fabricated
The data underpinning an infamous study of vitamin C for sepsis may be fraudulent, according to an analysis by an Australian physician and statistician that’s making waves among hospitalists and intensivists.
Kyle Sheldrick, MBBS, who is completing his PhD at the University of New South Wales, alleges that the pre- and post- comparison groups involved in the 94-patient study were too similar to be realistic.
“This is extreme,” Sheldrick told MedPage Today in an interview. “This is probably the most obviously fake data I have seen. … These groups are more similar than would be probable.”
The paper, led by Paul Marik, MD — a controversial figure during the COVID-19 pandemic who recently left his position at Eastern Virginia Medical School (EVMS) — has been the subject of much debate in the intensive care community since it was published in 2017. If clinicians could prevent death from sepsis with a simple regimen of hydrocortisone, ascorbic acid (vitamin C), and thiamine (the procedure was dubbed the HAT protocol), many lives easily could be saved.
“Obviously this paper sparked interest around the globe, but it’s hard to express how much excitement there was locally,” Bryan Carmody, MD, a pediatric nephrologist at EVMS, told MedPage Today. “I remember speaking to several faculty members who predicted that Dr. Marik might one day win the Nobel Prize.”
But many were skeptical from the beginning. “The effect size seemed just impossible,” said Nick Mark, MD, an ICU physician at Swedish Medical Center in Seattle. “It seemed too good to be true.”
Several larger and more methodologically robust studies followed, and to date, none of at least nine randomized trials — notably, the VITAMINS and VICTAS trials — has shown a similar reduction in mortality, raising suspicions even further, Mark said.
When Sheldrick posted his analysis earlier this week, it finally clicked:
“This was under our noses for 5 years,” Mark said. “This isn’t just a mistake. We know things can be done unethically, but to actually fake it? That it’s not just flawed, but perhaps actually fraudulent?”
Sheldrick told MedPage Today the key problem with the Marik paper is “probably the most common sign of fraud that we see, which is overly similar groups at baseline.”
Generally, when people are asked to fake a random distribution, they will consistently fake an even distribution, he explained: “You will fake data that’s far too close to the averages with not enough variation.”
Sheldrick said he first looked at the study methods, which noted a pre- and post- comparison design, rather than a randomized or matched case-control design. With such a design, one would expect a more random distribution of baseline characteristics, but that wasn’t the case for the Marik paper, he said.
He calculated P-values for the comparisons of the baseline characteristics using the Fisher Exact test, which he would expect to range fully from 0 to 1 given the study design. But he found the majority of P-values were 1 — meaning they were distributed perfectly evenly across two time periods — and only one fell below 0.5.
“We would expect to see these evenly spread between 0 and 1,” he told MedPage Today, with an overall average of around 0.5.
Sheldrick sent his findings to the journal CHEST and to Marik’s former employer Sentara Norfolk General Hospital, but hasn’t heard back from either as of press time.
Sentara Norfolk General Hospital did not return a MedPage Today request for comment as of press time.
A spokesperson for CHEST told MedPage Today it couldn’t confirm whether it has launched an investigation into the paper based on Sheldrick’s analysis. In an emailed statement, Nicki Augustyn, publisher with the American College of Chest Physicians, said, “When an allegation is formally submitted to the journal CHEST, the information provided is reviewed according to the guidance provided by the Committee on Publication Ethics. While an editor cannot comment on any specific allegation due to the confidential nature of the process, the journal CHEST takes ethical concerns very seriously and has an established, rigorous process for investigating all claims regarding articles we have published.”
Marik sent MedPage Today a statement through a spokesperson via email:
“The conclusions in my research have been validated in several meta-analysis [sic] and systematic reviews and show unequivocally that intravenous Vitamin C improves outcomes in sepsis patients. The allegations suggesting the data is fraudulent comes [sic] from a source with little qualification and in doing so is clearly demonstrating a lack of comprehension of scientific data analysis. The meta-analysis and systematic review published in Critical Care Medicine in 2021 stated ‘IV vitamin C monotherapy was associated with a significant reduction in overall mortality.’ A national cohort published earlier this year concluded that IV vitamin C for 5 days decreased mortality in hospitalized sepsis patients. I recommend your source examine this and other research on the data before making false allegations on social media. Such claims are harmful and do not add to the public discourse.”
This wouldn’t be the first time concerns have been raised about data in a paper that Marik co-authored. In November 2021, the Journal of Intensive Care Medicine (JICM) retracted an article by Marik and others on their MATH+ protocol for COVID-19.
The retraction followed a communication sent to the journal by Sentara Norfolk General Hospital that raised concerns about the accuracy of COVID-19 mortality data from the hospital used in the article.
According to an excerpt of Sentara’s communication published in the retraction notice, the hospital “conducted a careful review of our data for patients with COVID-19 from March 22, 2020 to July 20, 2020, which shows that among the 191 patients referenced in Table 2 that the mortality rate was 10.5%, rather than 6.1%. In addition, of those 191 patients, only 73 patients (38.2%) received at least 1 of the 4 MATH+ therapies, and their mortality rate was 24.7%. Only 25 of 191 patients (13.1%) received all 4 MATH+ therapies, and their mortality rate was 28%.”
“It seems a bit improbable for someone to discover two miracle cures in 3 years,” Mark told MedPage Today.
Mark noted that the 2017 paper has been widely cited. Even if the intervention hasn’t caused harm to patients directly, he noted that the resources invested in subsequent large, high-quality trials of vitamin C and sepsis could have been better spent.
“While I’m really glad we did high-quality studies and had brilliant people working on this, it’s kind of a shame,” he said. “Instead of studying vitamin C based on a faulty premise, we could have spent our efforts elsewhere.”