Quantifying Patient Investment in Novel Neurological Drug Development

Abstract

While the drug development literature provides numerous estimates of the financial costs to bring a new drug to market, the investment of patient-participants in the research process has not been described. Trial participants and their caregivers, like companies, invest time and undertake risk when they participate in prelicense trials. We determined the average number of patient-participants needed to develop a novel neurological drug. We created a cohort of 108 unapproved drugs first tested for efficacy between 2006 and 2011 and used ClinicalTrials.gov to capture enrollment in all subsequent prelicense trials of these drugs over a 9-year period. Our primary outcome was the average number of patients enrolled in prelicense neurological drug trials per drug that ultimately attained FDA approval, including patients who participated in both successful and unsuccessful development efforts. Five drugs (4.6%) were FDA approved, and 66,751 patient-participants were enrolled across successful and unsuccessful drug development efforts, resulting in an average of 13,350 patients for each drug attaining approval (95% CI 7155 to 54,954). Our estimates reveal the substantial amount patients and their caregivers contribute to private drug development.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13311-022-01259-y.

Introduction

The successful development of new drugs requires significant investments of money, time, and expertise. Private financial inputs to the clinical development pathway have been well characterized, with estimated costs averaging in excess of one billion dollars [1, 2]. While sponsors assume large financial risks in drug development, patient-participants in trials of unapproved drugs shoulder health risks and many additional burdens, such as extra research procedures and clinic visits. Patients also bear cognitive burdens entailed by a rigorous consent process and study compliance, which can be particularly challenging for neurology patients with cognitive impairment [3]. Although some of these burdens may be offset by therapeutic benefits, principles like clinical equipoise [4] and prior studies of risk/benefit in prelicense neurology trials [5] suggest that such offsetting benefits may be limited.

Quantifying the number of patients involved in the development of new neurological drugs renders more visible the extent to which members of the public shoulder some of the burden and risk of private drug development. This visibility can buttress arguments that patients and members of the public have strong moral claims to shaping policies on drug regulation and pricing. Such metrics can also be used to examine the interaction between research practices, regulatory policies, and patient welfare in research. For example, if drug companies are encouraged to overlook futility analyses and run phase 3 trials through to completion, or to launch phase 3 trials absent rigorous evaluation in phase 2 trials, greater numbers of patients may be exposed to ineffective drugs than if companies are encouraged to discontinue development of flagging interventions. In what follows, we estimate the number of patients with neurologic disease enrolled across the prelicense clinical trial landscape between 2006 and 2020 for each US Food and Drug Administration (FDA) approval of a novel drug. We include patient enrollment in trials of both agents that ultimately advance to approval and those that do not to capture the full patient investment per approval.

Methods

Our sample included neurology drugs and biologics first tested for efficacy between 2006 and 2011 inclusive, allowing for at least 9 years of follow-up based on piloting showing that most neurology drugs attain approval within this timeframe (see Supplement for details). We anchored our drug search at efficacy trials due to the inconsistent registration practices for phase 1 trials and a goal of focusing on patients and not healthy volunteers.

To begin, we searched ClinicalTrials.gov for all completed neurology trials initiated in this timeframe using a list of 357 neurology search terms generated by cross-referencing the “Nervous System Diseases” category on ClinicalTrials.gov with the International Classification of Diseases (ICD-11) chapter “Diseases of the nervous system” to focus on conditions that would be considered primarily neurological (see Supplement). Interventions were extracted and double-screened for inclusion based on whether at least one trial (a) enrolled patients with a neurological condition and (b) included a measure of efficacy as a primary or secondary endpoint. We defined conditions as neurological if they were categorized primarily under “Diseases of the nervous system” in the ICD-11; drugs for psychiatric, cardiovascular, metabolic, endocrine, and sleep disorders were thus excluded. In cases of ambiguity (e.g., cerebrovascular disease), we identified the drug target using DrugBank, AdisInsight, and peer-reviewed publications and included interventions that targeted a protein, receptor, or tissue found predominantly in the brain, spinal cord, or peripheral nerves. Disagreements were resolved by TSP and KK; further details are available in the Supplement. Interventions were excluded if they were (a) tested for efficacy in any indication prior to 2006, (b) FDA approved at the time of testing, or (c) not drugs or biologics. All eligible drugs were included (see eFigure 1 for screening flow diagram).

For each eligible drug, we identified synonyms on PubChem and searched ClinicalTrials.gov for trials initiated between the enrollment start date of the first efficacy trial and the earlier of either (a) the first FDA approval in a neurology indication (based on NDA filing date) or (b) 9-year follow-up. Trials of eligible drugs were excluded if they (a) enrolled healthy subjects or patients with non-neurological conditions or (b) were extensions of trials already included in our sample to avoid double-counting patients.

For each trial, study status, enrollment, phase, randomization, and enrollment start date were extracted automatically from ClinicalTrials.gov. Drug type, sponsorship (early involvement of large pharmaceutical companies or not), and orphan status of each indication were coded manually, in anticipation of performing a stratified analysis of whether any of these characteristics resulted in fewer patients per successful drug development effort. Enrollment was defined as “confirmed” if the trial’s completion status was marked as “Completed” or “Terminated” and if the enrollment type on ClinicalTrials.gov was listed as “Actual.” For trials with unconfirmed enrollment, we searched for publications to confirm the accuracy of the listed number of participants on ClinicalTrials.gov and updated our data accordingly. Trials were grouped into drug-indication pairing development efforts (“trajectories”) using the ICD-11 to define the indications (see Supplement). Patients from all eligible neurology trials initiated prior to FDA approval or the 9-year cut-off were included in the primary analysis. To describe the properties of drugs and trials captured by our sampling approach, we classified each trial according to whether it was randomized, enrolled children, and where it was located. Data were independently double-coded for quality control using a random sample of 10 drugs to confirm the replicability of the coding protocol; agreement was 94% or higher on all variables, so the remainder of the dataset was single-coded.

Our primary outcome was the average number of patients enrolled in prelicense trials per FDA-approved neurology drug, estimated by dividing the total patients enrolled in all trials of eligible drugs (regardless of subsequent approval) by the number of drugs that received FDA approval within 9 years. To establish a measure of dispersion in our estimate of patient enrollment per approved drug, we ran 10,000 simulations using our dataset, whereby individual drugs were randomly excluded and replaced with resampling to generate 95% confidence intervals. Secondary outcomes included the median patient enrollment across approved and unapproved drugs, the average patient enrollment per non-orphan drug approval, the number of trajectories launched per drug, and the relationship between order of indication tested and first indication receiving FDA approval.

We performed a sensitivity analysis at 12 years with all drugs with sufficient follow-up by July 2021. We also performed an analysis whereby patients enrolled in safety and pharmacodynamic studies preceding the first efficacy trial were included. All analyses were descriptive. Analyses were performed in R version 3.5.0. Outcomes were pre-specified and registered on Open Science Framework (https://osf.io/adfvp/); pre-specified analyses not reported here are available in the Supplement. This study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.

Results

Our sample included 108 drugs, five of which were approved within 9 years of first efficacy testing (4.6%) (Fig. 1, Table 1). None of the development trajectories for these five approved drugs were launched by large pharmaceutical companies. Most drugs were small molecules (66.7%), and a quarter received early support from large pharmaceutical companies (25.0%). Drugs were tested in 275 trials enrolling 66,751 patients, 53,105 of whom (79.6%) were enrolled in trials of drugs that did not advance to FDA approval. Development efforts consisted of 149 distinct drug-indication trajectories, 40 of which involved orphan indications (26.8%) (Table 2).

Fig. 1.

Neurological drug development process. In total, 66,751 patients contributed to the development of 108 drugs, 33 of which advanced to late phase testing (i.e., phase 3 trials, including phase 2/3) and 5 of which received FDA approval within 9 years. Color-coding represents the types of drugs at each stage of development, and the colored regions are proportional to the number of each type of drug that reached each stage. Drugs advancing to FDA approval in our cohort were peginterferon beta-1a (PLEGRIDY), valbenazine (INGREZZA), eteplirsen (EXONDYS 51), rimegepant (NURTEC ODT), and opicapone (ONGENTYS)

Table 1.

Approved drugs

Drug	Approval indication	Drug type	Years to approvala	Large company sponsorb	Patientsc	Trajectories
Peginterferon β-1a	RMS	Large molecule	4.0	No	1516	1
Valbenazine	Tardive dyskinesia	Small molecule	5.6	No	1072	2
Eteplirsen	DMD	Gene therapy	7.7	No	171	1
Rimegepant	Migraine	Small molecule	7.7	No	9326	2
Opicapone	Parkinson’s	Small molecule	8.2	No	1561	1

Characteristics of drugs in our sample that attained FDA approval for a neurology indication within 9 years of their first efficacy trial

^aYears from first efficacy trial enrollment launch date to NDA filing date

^bBased on the first efficacy trial sponsor. We did not account for drug acquisitions by large pharmaceutical companies later in development

^cTotal number of patients of any neurological indication, regardless of approval indication, enrolled in trials of each drug

DMD, Duchenne muscular dystrophy, RMS relapsing forms of multiple sclerosis

Table 2.

Sample characteristics

Drugs
	Total drugs		108
		Approved within 9 yearsa	5 (4.6%)
		Approved after 9 yearsa	3 (2.8%)
	Drug types
		Small molecule	72 (66.7%)
		Large molecule	12 (11.1%)
		Gene/cell therapy	14 (13.0%)
		Other/unclear	10 (9.3%)
	Sponsorshipb
		Large pharmaceutical company	27 (25.0%)
Trajectories
	Total trajectories		149
		Trajectories per drugc	1 (range: 1–5)
	Orphan statusd
		Orphan	40 (26.8%)
Trials
	Total trialse		275
		Trials per drugc	2 (range: 1–10)
		Trials per trajectoryc	1 (range: 1–8)
	Phasesf
		Phase 1	30 (10.9%)
		Phase 2	179 (65.1%)
		Phase 3	61 (22.2%)
		Otherg	5 (1.8%)
	Randomization
		Randomized	229 (83.3%)
	Trial location
		North America	101 (36.7%)
		Europe	60 (21.8%)
		Asia	28 (10.2%)
		Multiple	78 (28.4%)
		Other/unclear	8 (2.9%)
	Includes pediatric patients
		Yes	26 (9.5%)
Enrollment
	Total enrollment		66,751
		Patients per drugc	289.5 (range: 6–9326)
		Patients per trajectoryc	176 (range: 3–9266)
		Patients per trialc	110.5 (range: 1–2231)
	Enrollment statush
		Confirmed	62,761 (94.0%)
	Age
		Pediatric (< 18)	689 (1.0%)
		Adult (≥ 18)	65,216 (97.7%)
		Unclear	846 (1.3%)

Characteristics of drugs, drug-indication trajectories, trials, and patients included in our cohort. Enrollment, phase, and randomization were extracted from each trial record automatically; drug type, sponsorship, orphan status, trial site location, and pediatric inclusion were coded manually. Indication (for trajectory classification), drug type, and orphan status were double-coded for the full dataset, with discrepancies resolved by TSP and KK

^aApprovals for neurology indications only

^bSponsorship was defined based on the earliest efficacy trial of each drug. Large pharmaceutical company indicates a company in the top 10 pharmaceutical companies by revenue in the year of trial start according to Contract Pharma’s annual list; all other sponsors were either non-large companies or non-industry sponsors

^cValues are median (range)

^dOrphan status was based on the FDA definition of a condition with a prevalence under 200,000 patients in the USA. Orphanet, GARD, the FDA orphan designation database, and peer-reviewed publications were used to determine which conditions met this definition

^eOne trial (NCT01760005) tested two drugs head-to-head and was included in two trajectories but was counted only once for trial-level and patient-level variables

^fPhase 2 includes phase 1/2 trials; phase 3 includes phase 2/3 trials

^g “Other” includes four trials for which the phase was not specified and one trial of opicapone that was conducted outside the USA and registered as a phase 4 trial given that the drug was approved in the country of testing prior to FDA approval

^hEnrollment was considered unconfirmed if the trial was still recruiting, listed enrollment as “Anticipated,” or included a mix of neurological and healthy or non-neurological patients, and no publication or updated trial record was found to confirm final enrollment numbers

Our primary analysis showed that for each drug that advanced to FDA approval, an average of 13,350 patients participated in prelicense trials spanning the same time period (95% CI 7155 to 54,954). The median patient enrollment per drug was 290 (range 6–9326), with median enrollment for ultimately approved drugs being much higher than for drugs not attaining approval (1516 [range 171–9326] vs. 284 [6–5980]). Excluding orphan trajectories, 14,455 patients were required on average for each non-orphan drug approval (95% CI 7253 to 58,432).

Unlike in oncology where drugs are often tested across multiple indications [6, 7], few interventions were tested in more than one trajectory, with a median of one trajectory per drug (range 1–5). Of 108 drugs, 30 pursued two or more trajectories (27.8%), and all five approved drugs were approved for the first indication for which they were tested.

For the 62 drugs with 12 years of follow-up, total patient enrollment was 44,628, and the average number of patients needed to develop a new drug was 11,157 (95% CI 5175 to 52,927). The inclusion of 966 additional neurological disease patients enrolled in safety or pharmacokinetic studies preceding efficacy trials resulted in 13,543 patients to bring a drug to market (95% CI 7218 to 55,499).

Discussion

Approximately 13,000 patients participated in clinical trials for every new neurological drug receiving approval over the timespan studied. This estimate is similar to previous estimates for new oncology drugs (12,217 patients per drug) [8]. Our finding of a median enrollment of 1516 for drugs that ultimately attain approval is also in line with prior estimates of the volume of patients exposed to approved drugs in prelicense trials [9]. Four in five patients participating in prelicense trials were enrolled in trials where the study intervention did not advance to FDA approval; however, it is important to note that participation in unsuccessful development efforts often generates insights that advance future efforts [10]. That most unapproved drugs (91/103; 88.3%) were tested in fewer than 1000 patients before their development stalled (Fig. 2) suggests that the phased testing process spares many patient-participants from receiving ineffective or unsafe drugs in large, phase 3 trials.

Fig. 2.

Waterfall plot of patient enrollment per drug. Each bar represents the patient enrollment for a single drug across all trials and indications. Enrollment for drugs with negative outcomes, i.e., lack of FDA approval within 9 years of first efficacy trial, is shown on the positive y-axis, while enrollment for drugs that attained FDA approval within our timeframe is shown on the negative y-axis. Drugs are ordered by decreasing enrollment numbers

Among the five approved drugs, there was wide variation in prelicense enrollment. For instance, almost 10,000 patients were enrolled in prelicense trials of rimegepant for migraine, while less than 200 were enrolled for the orphan drug eteplirsen. That non-orphan drug development involved slightly more patients per approval may hint at greater efficiency of orphan drug development. However, our small sample of approved drugs limits our ability to compare per-patient gains for orphan disease vs. non-orphan disease. While orphan drug status certainly plays a role in these variations in prelicense enrollment, further study is needed to determine other characteristics of drugs and trials that render drug development more efficient and maximize the clinical impact generated by each patient-participant’s contribution. For instance, first-in-class drugs may require greater patient investment but produce greater clinical value than follow-on drugs, which may require less evaluation in phase 1 and 2 trials. Furthermore, the method of drug target selection in preclinical drug development may also influence the efficiency of clinical drug development. Our observation that drugs in our sample were tested in relatively few indications (median of 1 per drug) and that approvals occurred for the first indications tested suggests that in neurology, the selection of drug targets may be more disease-specific than in other domains, which may spare some patients from the potentially less favorable risk/benefit profiles of exploratory trials of drugs not specifically developed for their conditions.

Our analysis has several limitations. First, our sample of approved drugs was small, reflecting the challenging landscape of novel neurological drug development and a relative under-investment in this domain [11]. Our eligibility criteria also excluded several approved drugs that were first tested for efficacy outside of the 2006–2011 time window or that were previously approved for non-neurological indications (see eTable 1 for exclusion reasons). Our confidence intervals are thus wide. Second, some drugs may advance to FDA approval beyond our 9-year window. For instance, an additional three drugs in our sample were approved more than 9 years after first efficacy testing (eTable 2). However, our 12-year sensitivity analysis did not differ substantially from our primary outcome estimate. Lastly, as a cross-sectional study, estimates deriving from our historic sample may not reflect patient investment in drugs in development today, nor does it capture the contributions of patient-participants to the existing clinical research base that preceded the trials in our sample. Our estimates should thus be understood as providing a snapshot of the efficiency of neurological drug development landscape, which may vary over time and be influenced by the available knowledge base developed through decades of prior clinical research.

Our results underscore that drug companies and researchers are not alone in bearing the burdens of drug development. Trial participation is often framed as a net benefit to patients due to the prospect of both direct benefit through access to experimental therapy and indirect benefits such as financial compensation, extra medical attention, and a sense of purpose and contribution to science. Some patients may see these potential benefits as outweighing the risks and burdens of trial participation. However, this does not negate the fact that patient-participants and their caregivers commit a significant amount of time, effort, and welfare in supporting the development of new drugs [12]. Moreover, the principle of clinical equipoise logically entails that patients randomized to experimental treatment arms should not, on average, derive medical benefit over patients randomized to placebo or standard of care comparators, and they may in fact be harmed by receiving drugs that are less effective than the standard of care or that have more safety issues than the comparator. This has been shown empirically in neurodegenerative disease trials [5]. Drug development and research policy should thus strive to maximize the clinical value generated by the contributions of patients and their caregivers to the research enterprise by working to minimize redundant or uninformative research [13] and by reducing investment of patients in drug development trajectories that have little prospect of success. The latter might be accomplished by greater attentiveness to futility in phase 3 trials or proceeding to large phase 3 trials only after efficacy has been established in a phase 2 trial. Efficiencies in patient recruitment might also be achieved by deploying trial methodologies, like platform studies, that generate more data per participant [14, 15]. Fruits of this endeavor, including both the research findings and the treatments that are ultimately marketed, should be made accessible and distributed in a manner that reflects that drug development is a collaborative endeavor between large swaths of the public and the private sector.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contribution

Conception and design of the study: JK, NH, and AM. Acquisition and analysis of the data: AM, EG, CO, KK, and TSP. Drafting of the manuscript: AM and JK.

Declarations

Conflict of Interest

JK reports receiving consulting fees from Amylyx Pharmaceuticals. The authors otherwise declare no competing interests.