Editor’s note: This article was co-authored by Anders Sørensen.
In June 2014, Arif Khan et al. published a trial of 361 patients randomised to three groups: abrupt versus 1-week taper versus no discontinuation of desvenlafaxine.1 Curiously, although all patients had depression, the readers were told that data on depressive symptoms “are reported elsewhere.”
We were unable to find any such data when we searched on Khan and desvenlafaxine on PubMed. We were also unable to find the protocol for the trial and thought the trial had not been registered, as we did not find it by searching clinicaltrials.gov, and as neither the trial report, nor the PubMed abstract, mention a trial registry identifier.
The trial was funded by Pfizer, and we suspected the data on depression didn’t look good for the company since they had been omitted. Abrupt withdrawal of antidepressants can lead to an abstinence depression in many patients (e.g. an increase in the Hamilton score of at least 8 in a third of the patients treated with drugs with a short half-life).2
We tried to get the missing data from Khan, the corresponding author. Khan informed us that the trial had been registered (NCT01056289). The registry mentioned a secondary, so-called post hoc analysis, which we didn’t find in our searches, as Khan wasn’t one of the authors. This analysis was published in a little-known online journal,3 and there was a table showing depression scores assessed with the Quick Inventory of Depressive Symptomatology Self-Report. Very surprisingly, there were only small differences between the three groups although desvenlafaxine has a half-life of only 11 hours, which would have been expected to cause abstinence depression in many patients exposed to a cold turkey. This paper listed the trial identifier.
We were surprised that Kahn was not an author on the secondary publication, which only had present or former employees of Pfizer as authors. The figures in the online journal were totally unreadable, but we got them from the first author.
Khan informed us that, for the primary publication, he was given data tabulated by Pfizer employees and that he reviewed them and discussed how to present them. This suggests that the academic authors were not involved with data analysis.
We wonder how it can be possible that “Editorial/medical writing support was provided by Kathleen M. Dorries, PhD, of Peloton Advantage and was funded by Pfizer,” at the same time as the authors declare that all six of them “drafted the manuscript.1 Usually, only one person drafts a manuscript and if a medical writer paid by the drug company is involved, it is virtually always that person.
Khan wrote to us that the data are owned by a private company and not the US Government and that all the data are therefore proprietary. We strongly disagree. When patients volunteer for trials, which often involves running a personal risk—which is certainly the case here, where the trial was grossly unethical, as abrupt withdrawal increases the risk of suicide and violence4—the data cannot ethically be said to belong to a drug company. Drug companies have an ethical duty to make all their results public. Researchers should therefore never sign an agreement that the data are owned by a company. They belong to all of us. And researchers should not play hide and seek with the readers by omitting the trial identifier and publishing pivotal data “elsewhere.” All data that are relevant for the patients and their doctors should be published together.