At least two internal Amazon studies found a link between how quickly the online retailer’s warehouse workers perform tasks and workplace injuries, but the company rejected many safety recommendations out of concern the proposed changes might curb productivity, according to a U.S. Senate committee report.

The 160-page review issued this week was compiled by the Democratic majority staff of the Senate Committee on Health, Education, Labor and Pensions. The report is the result of a probe into Amazon’s warehouse safety practices that Sen. Bernie Sanders initiated last year.

The Vermont independent, a frequent critic of Amazon who chairs the panel, released an interim report in July that featured some findings from the investigation. The final report, which was mostly based on interviews with nearly 500 former and current Amazon workers, included more details, like the two internal studies and the reactions they received inside the company.

Amazon pushed back on the findings, saying in a blog post that Sanders “continues to mislead the American public” about the company’s safety practices and that the report was “wrong on the facts and features selective, outdated information that lacks context and isn’t grounded in reality.”

The Senate report Sunday said Amazon launched an internal study in 2021 to determine the maximum number of times a warehouse worker could perform the same physical tasks without increased risk of harm and potentially developing musculoskeletal disorders.

The team conducting the Amazon study, known as Project Elderwand, focused on workers who picked items from robotic shelf units. The study found the “likelihood of back injury increases” along with the number of items picked and identified an upper limit on repetitive movements — 1,940 — per 10-hour shift, the report said.

The study recommended using software to implement breaks “according to each worker’s rate.” It suggested expanding an existing Amazon program that recommended “microbreaks” and making them mandatory for employees who worked above the maximum pace.

The team stated that the success of a mini pilot program to test its idea would be conditional on “any negative impact to the (workers) or customer experience,” according to documents cited in the committee report.

Ultimately, Amazon did not make changes to reduce repetitive worker movements, the report said. The company told the Senate committee it chose not to do so due to “technical reasons” involving the proposed software program, the report said.

Amazon also said in its blog post that the Project Elderwand pilot program showed the suggested intervention was “ineffective.”

Amazon previously had undertaken another study, known as Project Soteria, in 2020 to identify risk factors for injuries and recommend policy changes that would improve worker safety. The multiteam initiative studied two policies Amazon implemented temporarily during the COVID-19 pandemic — giving workers more time off and pausing disciplinary measures “for workers who failed to meet speed requirements,” the report said.

The study found that both policies lowered injury risks and asked for their permanent adoption.

But company leaders denied the request, saying it might “negatively impact” productivity, according to Amazon documents in the Senate report. Amazon leaders also changed the focus of the Soteria study by telling the people doing the review to provide recommendations on how to improve productivity without worsening worker injuries, the report said.

Amazon disputed the report’s characterization of the events. “One team explored whether there’s a causal link between pace of work and injuries and another team evaluated the methodology and findings and determined they weren’t valid,” Amazon spokesperson Kelly Nantel said in a written statement.