Sample Identification
All original and supplemental novel drug applications were listed by the FDA on its publicly available FDA’s Drugs@FDA database [11]. Our sample included all SMKI cancer drugs, excluding nontherapeutic agents such as diagnostic and contrast agents. One investigator (Y.H.) manually extracted all application approvals of SMKI cancer drugs that occurred between 2001 and 2021 from the database. Initially, application approvals were identified by excluding the approvals that categorized by the FDA as “labeling revisions” or “manufacturing change or addition”. Then FDA’s letters accompanying the application approvals were examined to identify whether the approvals include a new indication. If one indication was related to two or more updated application approvals, such as “accelerated approval”, “new patient population” or “labeling change with clinical data”, we counted it as one indication.
Pivotal trials are the clinical trials that serve as the basis of the FDA approval. All pivotal trials that support both the original and supplemental applications were identified. A trial was considered to be pivotal if defined as such in the FDA medical review. If not specify, all efficacy trials that were described as essential to the application approvals were considered to be pivotal.
Data Extraction
Data were extracted by two investigators (Y.H. and L.M.). The following characteristics were collected for each application approvals: year of approval, the FDA’s expedited development and review programs or designations (approval pathway [accelerated, not accelerated], breakthrough therapy, fast-track, orphan drug, review type [priority, standard]), type of submission (original, supplemental).
For each application approvals, detailed information on indications was collected: cancer type (solid cancer, hematological malignancies), line of treatment (first, sequential), adjuvant treatment (yes, no), maintenance treatment (yes, no), type of administration (monotherapy, combination) and whether the indication was CDG biomarker-directed. An indication was considered to be directed by CDG biomarker if defined as such in the drug label, that is, the indication focused only on patients with the presence of a specific CDG biomarker as stated in the drug label.
Pivotal trials were identified from the FDA’s drug review dossiers and labeling. When more than one pivotal trial supported a single application approval, each trial was considered separately. That is, each approved indication could include one or more pivotal trials to support the application. For each indication, we then selected only one pivotal trial to collect the basic information of the trial and to compare between indications with and without CDG biomarker. The following steps were used to select the pivotal trials. First, we selected the trials that supported the initial application approvals of the indications, that is, trials that supported the postmarketing modifications of the indications were excluded. Second, the trial at the most advanced phase was selected. The collected basic information included: sample size, study design (RCT, non-RCT [SAT]), trial phase (phase I/II, III), type of blinding (double blind, open label), comparator(s) and primary efficacy endpoints. For trials with co-primary endpoints, the most definitive primary endpoint was identified, for example, overall survival (OS) prioritized over progression-free survival (PFS) and PFS over disease response rate.
Efficacy outcomes (including OS, PFS and disease response rate) were collected for all the pivotal trials that support the application approvals. When more than one pivotal trial supported a single application approval, each trial was considered separately and all the trials were selected to collect the efficacy data. If more than one efficacy data was reported for the same outcome in the same trial, most updated result was extracted. For trials assessing OS or PFS, we evaluated whether the efficacy data was clinically meaningful [30]. For all caner types, we considered OS gains of 2.5 months or more and PFS gains of 3 months or more to be clinically meaningful [31].
Statistical Analysis
We used descriptive statistics to characterize the indications and their supporting pivotal trials. To evaluate the number of approved indications over time, we estimated Poisson models with the number of indications as the dependent variable and a linear term for year of approval. The trends were estimated as incidence rate ratios (IRRs) with 95% confidence intervals (CIs). We also analyzed the difference in trends in the mean number of approved indications associated with CDG biomarker versus non-CDG biomarker directed therapeutics.
Indication and pivotal trial characteristics with and without CDG biomarker were compared by using nonparametric Mann-Whitney tests for continuous measures and using the Chi-Square tests for categorical variables. Associations between indication and trial characteristics and CDG biomarker were estimated as odds ratios (ORs) with 95% CIs using logistic regression with univariate model and multivariate models respectively. In the multivariate model 1, we adjusted for cancer type (solid cancer, hematological malignancies). To control potential confounding from type of approval, we additionally adjusted for approval pathway (accelerated, not accelerated) in the multivariate model 2.
Sensitivity analyses were performed to examine the robustness of the associations. First, we repeated our analysis excluding the indications in adjuvant or maintenance setting. Second, we repeated our analysis excluding the indications approved under combination therapy. A p-value less than 0.05 was considered significant. All the analyses were performed using SAS, version 9.4 (SAS Institute, Cary, NC, USA).