Publication bias and reporting bias are a serious threat to evaluate the level of evidence of new medicines and have been poorly studied for phase I trials. Such studies –whose main objective is to define the recommended dose– are however cornerstone for clinical development of oncology drugs. Our interest was therefore to evaluate publication bias and reporting bias on dose-escalation trials of protein kinase inhibitors (PKI) developed in oncology.

A systematic review was conducted between March 2020 and November 2021. Eligibility criteria were (i) PKI administered as monotherapy, (ii) healthy volunteers or patients with solid or hematological tumors, (iii) dose-escalation design. For each study, publications were searched on PubMed and Google Scholar. Publication lag was calculated as the time between the start date of the trial to the date of publication. Start date of the trial was used as a proxy because the end date of the trial is frequently lacking. Trials without publication were censored at the date of review. The quality of the reporting was assessed on the following variables: description of the (i) study design, (ii) dose levels, (iii) number of patients per dose level and (iv) number of dose-limiting toxicity (DLT) per dose level.

A total of 782 dose-escalation trials were identified of which 644 have a completed status. Among these trials, 413 (64%) were published with a median time of 6.49 years [6.07 – 6.84] after starting the trial. Ten years after, 29% [25% – 33%] trials are still unpublished. A post hoc analysis –not planned initially in our protocol– shows that the publication delay is shorter for trials studying a PKI with at least one marketing authorization. Ten years after, 20% [14% – 28%] trials studying a PKI marketed are still unpublished versus 32% [27% – 38%] for trials studying a PKI without marketing authorization. Among the 413 published trials, 57 (14%) does not specify the design used for the dose escalation. The dose levels are well reported in the articles, with very few missing data (n = 6 (1%)) adding that for 22 (5%) the reporting leaves some doubt about the order of these levels. The number of patients and DLT by dose level have missing data for 31 (8%) and 79 (19%) articles respectively.

The publication delay associated with the rate of unpublished trials and missing data are problematic for dose-escalation trials. There is an urgent necessity to standardize the reporting quality of dose-escalation trials.

Mots clés publication bias ; reporting bias ; dose escalation ; protein kinase inhibitors

Déclaration de liens d'intérêts Les auteurs n'ont pas précisé leurs éventuels liens d'intérêts