Skip to content
Lies, Damned Lies, and Statistics

Proton Pump Inhibitor Study Published With Questionable COVID-19 Data

American Journal of Gastroenterology published a seemingly groundbreaking study involving people who take so-called proton pump inhibitors are at a much higher risk of getting COVID-19. The study was heavily promoted before the data was questioned by several commentators.

Christopher Furlong / Staff / GettyImages

On July 7, 2020, the American Journal of Gastroenterology published a seemingly groundbreaking article claiming that in a study involving over 53,000 people, the ones who take so-called proton pump inhibitors — i.e., drugs like Nexium that treat heartburn and stomach pain — are at a much higher risk of getting COVID-19. This study was reported in the New York Times, Time, MSN, and others.

Oddly enough, all of the data came from a survey firm, not from hospital or medical records. That is, people were given a survey about themselves and their medical conditions, with no independent verification or administrative records involved.


As has been pointed out by a number of Twitter commentators, the initial results should have made the researchers question their data, if not their entire ability to conceive of a research project in the first place. Table 1 of the study claims that while the overall survey respondents were a nationally representative sample — fairly balanced across age, gender, income, race/ethnicity, etc. — there were some surprising imbalances among the survey respondents who got COVID-19:

  • 74.5% were between 30 and 39 years old;
  • 64.7% were female;
  • 69.7% were Latina;
  • 69.6% had only a high school education;
  • 63.5% had an income over $200,000 a year;
  • 73% were daily smokers; and
  • 68.5% lived in the South.

Does anyone believe that in a nationally representative sample of Americans, a huge majority of the COVID-19 cases would be 30-something Latinas who live in the South, smoke every day, have a high school education, but nonetheless make over $200,000 a year? This study shows why the number one rule for researchers should be to take stock of whether the underlying data are believable.

Grants

Arnold Ventures funds projects to understand problems and identify policy solutions.

Map of the U.S. made of Arnold Ventures icons