Reacción a "Use of GLP-1 analogue drugs is associated with benefits but also health risks"

The paper describes a data-driven approach to studying possible beneficial and adverse effects of GLP-1 agonists like semaglutide (Ozempic), which are widely used to treat diabetes and obesity. The sophisticated statistical data techniques employed by the researchers are like those used to study effects of human genetic variations. In genomic analysis the data come from genome sequencing, but here they come from routine diagnostic, treatment and administrative records of millions of patients.

The researchers compare event rates in users of GLP-1 agonists with those observed in patients selected for other diabetes treatments. Rather than testing prior hypotheses about benefits and harms of these new medicines the analysts look for patterns in the data that might signal previously unsuspected effects. Some of the beneficial effects seen in their data, particularly in neuropsychiatric and substance abuse disorders, are supported by other studies. Some may be related to improvements in control of weight and metabolism, well known effects of the drugs. Others may be due to the play of chance or bias, because of differences between patient groups who are chosen to start different diabetes medications.

This is an observational study, not a randomised trial, and the authors caution against basing treatment recommendations on these data, without further confirmation. Such reticence is justified. Flawed observational studies of another diabetes drug, metformin, concluded erroneously that the drug prevented cancer. Analysis of data from randomised trials disproved that theory. This type of ‘big data’ research will generate lots of statistical associations, some of which will be spurious. It is not yet clear whether these data driven methods for exploring beneficial and harmful drug effects are a genuine advance over traditional inferential approaches.