AMSTERDAM – The attempt to determine whether biologics such as tumor necrosis factor inhibitors (TNFi) inhibit progression of ankylosing spondylitis has been pursued with observational studies, but these types of studies will never definitively answer the question, according to Robert B.M. Landewé, MD, PhD, professor of rheumatology at the University of Amsterdam.
“The methodology is sensitive to a lot of measurement error, making the results spurious,” Dr. Landewé said in an interview, recapping remarks he made in a presentation at the European Congress of Rheumatology.
This was disappointing to many investigators, including several speaking in the same symposium where Dr. Landewé made his remarks. Randomized, controlled trials that employ serial radiographs to document changes in ankylosing spondylitis are expensive, making observational studies an attractive surrogate, but Dr. Landewé said such studies are associated with an inherent risk of residual confounding.
In addition, he believes the effect size of biologics on progression, if it exists at all, is likely to be subtle. In the observational studies that have concluded that there is protection, complicated statistical analyses have been typically employed to produce a significant finding.
Observational studies do have hypothesis-generating value, according to Dr. Landewé, but he cautioned that they produce “more questions than answers.” He also emphasized that the inflammation-related progression that leads to bone growth in ankylosing spondylitis is different than it is in the destructive inflammatory diseases, such as rheumatoid arthritis, where the issue is bone loss.
It is rational to assume that effective anti-inflammatory therapy would prevent progression of inflammatory diseases, but Dr. Landewé said in his presentation that this is the type of bias that undermines the value of observational studies for reaching objective conclusions. Unlike the results of a registered randomized, controlled trial, which will be known to be consistent or not with the underlying hypothesis, there is a strong risk that data in an observational study will be reworked until they produce the desired result.