Costs and Causes of Oncology Drug Attrition With the Example of Insulin-Like Growth Factor-1 Receptor Inhibitors

Key Points Question What are the associated expenses of clinical research and what factors underly the translational failure of inhibitors of the insulin-like growth factor-1 receptor (IGF-1R) in oncology? Findings In this cross-sectional study, 16 inhibitors of IGF-1R underwent 183 clinical trials in more than 12 000 patients; none of the agents was approved for clinical use in oncology practice and the trials were estimated to have had expenses of greater than $1.6 billion. Half of the published in vivo preclinical data analyzed showed less than a 50% inhibition of tumor growth by IGF-1R inhibitors. Meaning With high attrition rates for oncology drugs, the fruitless and expensive clinical trials of 16 IGF-1R inhibitors draw attention to the need for improved preclinical models and better decision-making before trials are launched, reducing substantial financial losses and avoiding exposure of patients to potential toxic effects.


Introduction
Over the past 2 decades, the pharmaceutical industry has become heavily focused on the discovery of therapeutic agents for the treatment of metastatic cancer. 1 These agents largely either target genetic changes implicated in cancer pathology to provide targeted therapies 2 or they modulate the immune system. 3 However, development of oncology drugs has been reported to have a persistent attrition rate of greater than 95%, [4][5][6] highlighting the disparity between positively assessed preclinical drug activity and subsequent inactivity in patients. The costs of clinical trials in oncology have been estimated to exceed those of other therapeutic areas so that failure is likely to be expensive. 7 Failure is not only expensive in time and money but is disappointing for the scientists and clinicians dedicated to projects that flounder. Failure is ultimately most disappointing for patients, some of whom may be exposed in clinical trials to therapeutically inactive drugs that carry toxicity.
To our knowledge, estimates of the costs of clinical drug attrition in oncology have not been published. We have chosen to estimate the development expenses associated with the search for clinically active, targeted inhibitors of the insulin-like growth factor-1 receptor (IGF-1R). [8][9][10][11][12] Insulin-like growth factor-1 receptor inhibitors (ie, small molecules, antibodies, and cell therapy), were chosen as an example as we were able to capture the results for 16 IGF-1R inhibitors evaluated in trials against a broad range of tumor types by different pharmaceutical and biotechnology companies, many of which had published preclinical data. All the inhibitors, alone or in drug combinations, failed to demonstrate clinical activity deemed sufficient for approval in oncology practice.
We have estimated the expense of the clinical phases of IGF-1R inhibitor development programs, collating details of treatments, including the numbers of patients entered in trials. We discuss the quality of the preclinical data leading to the launch of IGF-1R inhibitors into clinical trials and address more broadly the financial consequences of failure. We attempt to place the failure of the IGF-1R inhibitor programs into a temporal scientific and clinical context.

Creation of Databases of Clinical Trials of IGF-1R Inhibitors
This study of published research data did not involve personal medical records and does not constitute human participant research. We searched public databases (Google Scholar, PubMed, SCOPUS, and Web of Science) to find details on inhibitors of IGF-1R. Specifically, the key words used were molecular IGF-1R, therap*, targeted therap*, cancer, oncology, clinical trial, and drug development.
A database of clinical trials of IGF-1R inhibitors was then created by searching the ClinicalTrials.gov registry (NCT) 13 for the key terms IGF-1R and the individual drug names or codes, with the date range from January 1, 2000, to July 31, 2021. Clinical trials not in the field of oncology, identified by pathology, were removed from the database. Trials were annotated for trial title, drug code, trial sponsor (industry or other), cancer indication, date of trial, patient numbers, trial phase, and program costs. Consideration of the study lead organizations allowed an insight into the primary sources of financing for each clinical trial. As such, organizations were considered to belong to the pharmaceutical or biotechnology industry, academia, the US National Cancer Institute, or other.

Estimate of Clinical Development Expenses
The first research and development (R&D) expense estimates were from a proprietary database made accessible to us by Evaluate Ltd, 14 which provided data for 129 of the total of 183 trials found.
United States publicly traded companies are required by the Securities and Exchange Commission to file a detailed annual report, known as a 10-K. For many small and medium-sized biotechnology and pharmaceutical companies, the 10-K, and sometimes other publicly available documents, contains R&D expense information at the level of individual drugs. Evaluate Ltd collates this drug-level expense information on a nominal basis (ie, not inflated or deflated to a reference year) and combines it with other data in the public domain (eg, from ClinicalTrials.gov and other company disclosures in the 10-K) to derive per-patient benchmarks. The empirical phase 2 and 3 trial benchmarks consider a fixed trial expense plus a per-patient expense. There are further adjustments for geographic location and trial duration vs the technology-by-EPHMRA benchmark. The benchmark estimates are checked by comparing forecast R&D expenses with company-reported R&D expenses for the top 20 global biopharmaceutical firms. Evaluate Ltd method centers on what can be thought of as an accounting view of R&D expenses since they are derived from accounting-based reports (eg, 10-K filings). The Evaluate Ltd figures do not include the cost of capital that one would see in an investment view of R&D costs. However, the figures in the 10-K filings will generally include program expenses above the narrow cost of clinical trials (eg, manufacture of drugs for trials, any ongoing nonhuman toxicology studies, data processing, and preparing for regulatory submissions).
Furthermore, companies will vary in what they exclude and include in when they report drug-level R&D expenses in their 10-Ks. We note that the estimates of Evaluate Ltd are used extensively in the drug industry by companies that know their own R&D costs. Details and a general primer on R&D cost estimates are provided in the eMethods in Supplement 1.

Estimate of Expenditure
We found 54 clinical trials of IGF-1R inhibitors for which Evaluate Ltd did not provide an expense estimate. To estimate these expenses, the mean cost per patient per trial phase was calculated from the Evaluate Ltd data given in eTable 1 in Supplement 2 and eTable 3 in Supplement 2) and multiplied by the number of patients described in the NCT database 13 for each of these 54 trials. Eleven of these trials were performed by industry or biotechnology companies and we assumed their costs at the mean (per phase) of trial expenses estimated by Evaluate Ltd. Since estimates of the costs of academic and other trials, such as those by the US National Cancer Institute, are inaccessible, we estimated a range of possible expenses at 1 × the Evaluate Ltd mean, 0.5 × the Evaluate Ltd mean, and 0.2 × the Evaluate Ltd mean (eMethods in Supplement 1).

Basket Trials
When the IGF-1R inhibitors were in used in basket trials (ie, IGF-1R inhibitors were only one of several different interventions), we first took the expense of the complete basket trial and divided this by the total number of patients enrolled to determine the expense per patient. We then investigated the numbers of patients enrolled in the IGF-1R inhibitor arm from available publications and estimated the expense of that arm alone by multiplying the cost per patient by the number of patients in the IGF-1R inhibitor arm.

Analysis of Selected In Vivo Preclinical Data
A search for published articles describing preclinical data in vivo for each inhibitor was made using PubMed, using the code number or drug name reported in Table 1 (with one drug for oncology only, subsequently developed for Graves disease by Horizon 15 ). The percentage tumor growth inhibition was calculated as described in Carboni et al 16

Clinical Trial Data and Program Expenditures
We found 16 IGF-1R inhibitors, small molecules, a mixed antisense cell therapy, and antibodies that entered oncology clinical trials between 2003 and 2021 (Table 1). Table 1 also presents our estimate of the number of patients who were entered into trial programs as found in the NCT database 13 and, if not, the mean patient number was used for each phase. We estimate that a total of 12 396 patients were entered into 183 trial programs. Details are presented in eTable 1 and eTable 2 in Supplement 2, with basket trials reported separately.     Table 2 presents a survey of the published preclinical in vivo data testing IGF-1R inhibitors, detailing the inhibitor used, the publication details, the xenografted tumor type, brief observations on the assay methods, and results, with estimates of percentage tumor growth inhibition for 9 inhibitors tested as a single agent against a broad range of tumor xenografts.  Of the 62 cell line results annotated, 31 had a percentage of tumor growth inhibition of less than 50%.  treatment of Graves disease. 15 More than 12 000 patients were estimated to have entered these futile trials. Used as single agents, the drugs had an acceptable toxicity profile, although some drug combinations were toxic. 51,52 For example, in the study of ganitumab with hormonal treatment of receptor-positive breast cancer, serious adverse events developed in 25% of the patients in the ganitumab group compared with a placebo, with the most common grade 3 or higher adverse event being neutropenia. 51 Publications analyzing these trials addressed important issues that could have explained the lack of clinical efficacy of the IGF-1R inhibitors. 53-59 They included redundancy within the IGF-1Rstimulated signaling pathways, with compensation of signaling by alternative pathways-an occurrence common to most signaling inhibitors 60   Specification of the category of others is included in eTables 1 and 2 in Supplement 2 included adrenocortical tumors, brain tumors, carcinomas (not specified), chronic myeloid leukemia, gastroesophageal tumors, head and neck tumors, melanoma, mesothelioma, multiple indications, multiple myeloma, neoplasms (not specified), neuroendocrine, nonhematologic, ovarian, solid tumors (not specified), thymus, and not specified.   (Table 2), despite overall preclinical in vivo data having been described as compelling. 11

Discussion
Xenograft data considered to be predictive of the successful clinical activity of IGF-1R inhibitors, all of which subsequently failed to show clinical efficacy, can either be considered to invalidate the models themselves (when there was 100% tumor growth inhibition but no clinical activity) or, at the least, the interpretation of them. In 2014, drug researchers at AstraZeneca articulated a 5-dimensional framework that reduced their drug attrition in clinical trials: more rigorous target validation and improved preclinical models were key aspects that required attention. 65 This study of IGF-1R inhibitors supports that recommendation, as does the recent suggestion by some of us to use more effective and stringent decision tools during preclinical development. 63 Published estimates of drug development costs generally focus on successful drug registrations, whereas herein we estimate the cost of failure (Figure 1). The 183 IGF-1R trials had cumulative R&D expenses estimated to be greater than $1.63 billion over the period of our survey. Looking at the drug industry, cancer R&D that is failing or destined to fail will currently incur an annual expense that is in the order of $50 billion to $60 billion (eMethods in Supplement 1). have been a force creating strong expectations. 68,69 It is also possible that company management, with these high expectations, faced with strong competition and the need to fill their drug pipelines, proceeded in a case of herd instinct, as discussed recently. 70 To quote Borup et al 71 in an analysis of expectations in scientific research, "behavior is not only based on rational risk-return considerations, but also influenced by expectations and perceptions of other's behavior." It will be interesting in the future to analyze the costs and outcomes of the current intense competition in oncology, with many me-too programs.

Moser and Verdin
What might be learned from the failure of the IGF-1R inhibitor programs? We see at least 3 lessons. The first relates to the evidential hurdles that cancer drugs should pass before moving into the clinic; specifically, the predictive validity of the preclinical models and performance standards that drug candidates should meet in those models. 63 The second relates to technical diversification to avoid overinvestment in some mechanisms (eg, IGF-1R) and underinvestment in others. 66,67 The third relates to a lack of rigorous analyses of major translational failures. 63

Limitations
This study has limitations. The limitations in estimating drug development costs and failure are elaborated in the eMethods in Supplement 1. In 129 of the trials we found of IGF-1R inhibitors, the Evaluate Ltd algorithm (eMethods in Supplement 1) was used to estimate expenditure on these trials.
This algorithm could be challenged, although many international pharmaceutical companies use data generated by it. When we lacked data from Evaluate Ltd, we used their mean expenditure per patient per phase of clinical trial. It is not based on real data that were gleaned from company reports and is therefore questionable. Trials performed by academic centers or institutions, such as the US National Cancer Institute, do not have transparent sources to permit estimations of trial expenditure and so had to be estimated.
We have discussed herein clinical trials and preclinical data that date from the first 2 decades of 2000. It is possible that discovery and improved use of biomarkers for IGF-1R inhibitors could reduce these examples of clinical trial failure, although we are not aware of recent data to substantiate this.
We may have overlooked some of the publications of in vivo preclinical data that allowed us to derive the percentage inhibition of tumor growth by IGF-1R inhibitors. It is also likely that more data, especially from pharmaceutical companies, were not published and that only representative data or optimal data on in vivo assays were published. However, we were able to capture data from 35 publications in which a wide range of cancers xenografted in mice were used. We have not analyzed the data from in vitro tests of IGF-1R inhibitors on cell lines as it is likely that the in vivo data were most important in the decision to progress a candidate drug to clinical trial.

Conclusions
We do not dismiss the challenges of drug discovery in cancer. During the period that IGF-1R inhibitor projects were launched (1999-2009), it was reported that 83% of the claims that cancer biology could be successfully translated into treatments proved futile, and of drugs that offered an overall survival benefit, it was for a mean of only 6 months. 72 This limited impact of many systemic therapies to improve overall survival from cancer has continued. 73,74 The biology of cancer is complex and understanding of cancer mechanisms continues to evolve. In the early years of the 2000s, when IGF-1R inhibitor programs were being launched, 6 hallmarks of cancer were identified; in 2022 this had grown to 14. 75 Our growing understanding of the complexity and heterogeneity of cancer should provide pause for thought about the human and financial resources involved in the enterprise of drug discovery and to where effort and resources may be better focused to reduce cancer mortality.