Justifying adjustment
Speaking during the audience question period Giuseppe Del Priore, MD, from Morehouse School of Medicine in Atlanta, Georgia, said he preferred unadjusted results when examining real-world data, “because that’s the benefit,” and he questioned why the researchers had adjusted their numbers.
Dr. Vernieri explained that the adjustments were made to account for the important imbalances in the baseline characteristics of the patients.
“When we plotted unadjusted curves, we did not find statistically significant differences between these three drugs, only a trend toward the direction that I showed you today,” he said. “However, as you saw from the tables showing the characteristics of patients, there were important imbalances in terms of important prognostic factors in the three patient cohorts. So, I think that, for this kind of data and based on this level of imbalance, adjustment is necessary.
“To reinforce our conclusions, what we did was also to perform a propensity score match–based analysis,” Dr. Vernieri continued. “I did not have the time to show the results today, but these data were fully in line with the study conclusions. And we also performed a backward selection of variables. So, we basically selected variables more likely to be associated with patient prognosis. And also those models confirm the study conclusion. So I think the conclusions are quite solid.”
Dr. Del Priore, an adjunct professor of obstetrics and gynecology with a specialty in oncology, on the other hand, said he was not convinced that any of the drugs might be better or worse in the actual population treated.
“I still maintain that unadjusted real-world data should be presented and then only a limited adjusted analysis performed using the most unbalanced variables,” he said. “To do more elaborate adjustments may falsely imply a difference in drug choice and outcomes which never should be the conclusion with observational studies. Instead, the conclusions should be that, with typical use, the following similarities in PFS and OS were observed. Then point out how drug choice and important prognostic variables might be linked, thus limiting the generalizable conclusions even further.
“I would conclude that prospective studies should balance for the variables used in the PALMARES-2 analyses, which actually may have been chosen for adjustment post hoc,” Dr. Del Priore said.
The study was funded by the Italian Association for Cancer Research, the European Research Council, the Ministero della Salute, the Scientific Directorate of Fondazione IRCCS Istituto Nazionale dei Tumori, Giuliani’s Foundation and Roche. Dr. Vernieri reported consulting or advisory roles with Daiichi Sankyo/Astra Zeneca, Novartis, and Pfizer; speakers’ bureau roles with Accademia Nazionale Di Medicina (ACCMED), Istituto Gentili, Lilly and Novartis; and research funding from Roche. Dr. Del Priore reported no conflicts of interest and disclosed that he is chief medical officer at BriaCell.