Patient Participation in Clinical Trials of Oncology Drugs and Biologics Preceding Approval by the US Food and Drug Administration

Key Points Question How many patient study participants are needed to obtain a first US Food and Drug Administration approval for a new anticancer drug or biologic therapy? Findings In this cohort study of 120 drugs and biologic therapies, more than 12 000 patients participated in prelicense clinical trials for every new drug or biologic approved by the US Food and Drug Administration. When confined to drugs and biologic interventions with intermediate to substantial clinical impact, nearly 40 000 patients were required per approval. Meaning These results indicate that in addition to involving large private expenses, oncology drug and biologic development entails a large subsidy of altruism, time, and welfare from patients themselves.

This supplemental material has been provided by the authors to give readers additional information about their work.

eMethods 1. Protocol Modifications
Modifications to secondary outcomes: -We categorized drugs into four classes (immunotherapy, targeted, cytotoxic and other), rather than classifying drugs as small or large molecules, as the former was considered more descriptive (Modified June 2019).
-Early enrichment was added as an extraction item and for inferential testing (Modified June 2019).
-We had multiple drug approvals in our cohort and this allowed us to estimate and compare the number of patients needed to develop a new drug in novel/not novel, early enriched/not early enriched, immunotherapy/not immunotherapy, large pharmaceutical company/not large pharmaceutical company and early launch (2006-2008)/late launch (2009-2010) categories, as secondary outcomes. This was felt to provide an accurate assessment of drug development efficiency, as it relates to the patient burden of clinical testing. It therefore replaced our original intention to compare mean patient enrollment by drug category (Modified December 2019).
-Given large variation in patient enrollment figures, we reported median patient enrollment per drug and interquartile range, rather than mean patient enrollment per drug. We also reported median number of trials per drug and interquartile range.
-We performed sensitivity analyses on our primary outcome for drugs that had 10 and 12 years of follow-up (Modified October 2019; performed in January 2020 to allow for data up to 2019-12-31 to be included). We added a further sensitivity analysis to estimate total pre-license enrollment from Phase 1 divided by the number of FDA approvals (Modified January 2020).
-Based on reviewer comments, we performed a post-hoc assessment of orphan drug status, and compared the number of patients per FDA approval in orphan versus non-orphan designated drugs (Modified March 2021). Drug and biologic interventions were included in our cohort based on the following criteria: We excluded the following drugs and biologic interventions from our cohort: 1. Advanced into first efficacy oncology trials before 2006 2. Received FDA approval in oncology prior to first registered efficacy trial 3. Received FDA approval in a non-oncology indication within 8 years of the first identified oncology efficacy trial (to exclude drugs repurposed for an oncology indication) 4. Aimed at primary prevention of cancer or symptoms secondary to cancer 5. Drugs or biologics that had no trials with a site in the United States 6. Devices 7. Cells, Viruses, Plasmids (generally lacking trade names in registration documents, making tracking impossible)

. Assessment of Drug Class
Using trial registration records and drug descriptions from NCI thesaurus (https://ncithesaurus.nci.nih.gov/ncitbrowser/), with additional search of Drugbank (https://www.drugbank.ca) and PubMed (https://www.ncbi.nlm.nih.gov/pubmed/) using drug name and synonyms as required, each drug in the cohort was categorized in one of four drug classes (immunotherapy, targeted, cytotoxic or other). A permissive definition of immunotherapy was used, such that evidence of any manipulation or stimulation of the immune system to recognize and/or target cancer cells, in addition to the direct targeting of immune cells, was classified as an immunotherapy drug. Examples of immunotherapy drugs include CTLA-4 inhibitors and PD-L1 inhibitors. Targeted drugs inhibit/activate specific molecular targets, such as tyrosine kinase receptors or enzyme poly ADP ribose polymerase (PARP). Cytotoxic drugs affect all dividing cells, leading to cell death. Examples include topoisomerase II inhibitors, antimetabolites and alkylating agents. Our model was hierarchical, such that drugs with characteristics of more than one class were characterized based on their more innovative mechanism. Immunotherapy was considered the most innovative, followed by targeted therapy and cytotoxic therapy. Finally, drugs not fitting into any one of these three categories were labeled as other. For example, hormone therapy was classified as other in our study.
Duplicate assessment of drug class was carried out by two evaluators (NH & SZ) with any disagreements adjudicated by a third (JK). The first thirty drugs assessed were considered teaching cases, to ensure adequate agreement between the two primary assessors. The following ninety drugs were assessed independently, without discussion. Agreement between the evaluators using an unweighted Cohen's kappa was 0.808.

eMethods 4. Early Enrichment Assessment
Our definition of enrichment was based on the definition of personalized therapy by Schwaerderle 2015 1 , in which we considered a patient population to be enriched if either i) a biomarker was used in patient selection; or, ii) no biomarker was used, but a specific patient population was selected, and at least 50% of patients with the particular disease entity are known to possess a specific biomarker. We considered a drug development program in our cohort to employ an early enrichment trial design if, in one of its first two oncology efficacy trials, the trial registration record indicated selection of an enriched patient population, and the mechanism of action of the drug targeted the specific signalling pathway (directly or one stepped removed) for which the patient population was enriched. In this way, our definition of enrichment mirrored the FDA's description of "predictive enrichment" in which a protein or genetic marker related to the drug's mechanism of action is used to identify the treatment population. 2 Duplicate enrichment assessment was carried out by two evaluators (NH & SZ) with any disagreements adjudicated by a third (JK). The first 30 drugs assessed were considered teaching cases, to ensure adequate agreement between the two primary assessors. The following 90 drugs were assessed independently, without discussion between assessors. Agreement between the evaluators using an unweighted Cohen's kappa was 0.681.The following steps were followed in our enrichment assessment: 1. Identify the first two oncology efficacy trials for each drug using the Clinical Trials Viewer (https://trials.bgcarlisle.com/) with a full synonym list.
2. For each of the two identified trials, read the title, introduction and inclusion/exclusion criteria of their registration records on clinicaltrials.gov to determine if the patient population was enriched. Only a portion of the patient population needs to be enriched to fulfill this criterion.
3. If there is an enriched patient population, use NCI thesaurus (https://ncit.nci.nih.gov/ncitbrowser/) to determine if the drug is enriched by determining if the mechanism of action matches the selected patient population.

eMethods 5. Novelty Assessment
Our novelty assessment was based on the premise that there is significant industry awareness when a pre-license drug reaches Phase 3 testing, prompting competitors to initiate clinical evaluation in similar molecules. These latter molecules, if they mirrored the type and mechanism of action of the original drug, were not considered novel. By mechanism of action we referred to the target (e.g. enzyme or receptor) of a drug which resulted in the anti-neoplastic effect. A drug in our cohort was only considered novel if one of the following conditions were met: i) no other drug with a mechanism of action that covered the main mechanism of action of the drug in our cohort was found; ii) there was no other drug of the same type (e.g. monoclonal antibody) that covered the main mechanism of action of the drug in our cohort; and, iii) drugs of the same type and with the same main mechanism of action reached Phase 3 testing (or FDA approval if no Phase 3 was conducted) only after the start date of the first oncology efficacy trial of the drug in our cohort. If none of the above criteria were met the drug in our cohort was considered not novel. If there was not enough information to determine novelty status, then a category of nonapplicable was used.
Our criterion for novelty was similar to the first-in-class drug category described by Lanthier et al. 3 However, given our evaluation of a pre-license cohort of drugs and biological interventions, we used first intervention to launch Phase 3 testing, rather than first to gain regulatory approval, in our assessment.
In the following visual example, drugs X, Y and Z all have the same mechanism of action and are of the same type. Drug X is the drug in our cohort, and is considered not novel because Drug Z initiated Phase 3 testing prior to the launch of the first oncology efficacy trial of Drug X. Although Drug Y also reached Phase 3 testing, the date of launch of Drug Y's Phase 3 trial occurred after the launch of Drug X's first oncology efficacy trial. Therefore, it is Drug Z and not Drug Y than renders Drug X not novel.
Duplicate novelty assessment was carried out by two evaluators (RB & EG) with any disagreements adjudicated by a third (NH). The first twenty drugs assessed were considered teaching cases, to ensure adequate agreement between the two primary assessors. The following one hundred drugs were assessed independently, without discussion between assessors. Agreement between the evaluators using an unweighted Cohen's kappa was 0.738.
The following steps were followed to assess novelty: 1. Search NCI thesaurus (https://ncit.nci.nih.gov/ncitbrowser/) for the mechanism of action of the drug in our cohort and for its synonyms. (If no entry on NCI thesaurus, or mechanism of action unclear, also check PubChem (https://pubchem.ncbi.nlm.nih.gov/search)). -If there is an article which provides a review of the current drugs under investigation for a specific type of drug, then that can also be used as a useful source to identify drugs with similar mechanism of action.
6. For each new comparator drug identified, check NCI thesaurus for mechanism of action, drug type and drug synonyms. If no NCI thesaurus entry identified, then also check PubChem.
7. If similar mechanism of action and drug type, then used the Clinical Trials Viewer (https://trials.bgcarlisle.com/) with full synonym list of the comparator drug (using "OR") to determine the start date of its first Phase 3 and note down the date.
8. If similar mechanism of action and drug type, and if significant evidence of clinical trialing in the comparator drug such that FDA approval is reasonably foreseeable, using drug name and synonyms for the comparator drug, evaluate FDA approval status and date of approval by searching Drugs@FDA (https://www.accessdata.fda.gov/scripts/cder/daf/) and FDA's Biological Approvals by Year (https://www.fda.gov/vaccines-blood-biologics/development-approvalprocess-cber/biological-approvals-year).

eMethods 7. Orphan Drug Classification
As a post-hoc analysis, we assessed orphan drug designation status for all of the drugs and biologic interventions in our cohort. Using drug names and synonyms we searched https://www.accessdata.fda.gov/scripts/opdlisting/oopd/ for presence of FDA orphan designations. We classified a drug or biologic as having gained orphan status if it had ever received an orphan designation for any indication in a trial in our cohort, which was not subsequently withdrawn.