5.1. Phase Durations

The phase duration parameter refers to the time it takes to complete a given stage of development depicted in Figure 1. For the non-clinical stage, our estimate represents the time it takes from synthesis of the compound to the start of human trials, which includes early exploratory research for target discovery, hit generation and target identification; lead optimization; preclinical work involving animal testing to develop dosing and toxicity models; and obtaining an IND approval from FDA to begin testing in human subjects. We used published studies and information compiled from FDA’s DASH database to estimate average phase durations across all development stages (Table 3) by therapeutic area. From Table 3, Phase 3 is the longest (38.0 months) drug development stage across all therapeutic areas followed by post-approval Phase 4 (36.6 months), Phase 2 (34.0 months), non-clinical stage (31.2 months), and Phase 1 (27.8 months). The average time for the FDA review phase is 16.2 months. This includes the time the sponsor spends on responding to any questions and/or information requests from the FDA as well as preparing major/minor amendments, if needed. Thus, the estimate does not solely reflect the time FDA spends on reviewing the application. While there is variation in phase durations across the different therapeutic areas, this ranking is generally stable with Phase 3 comprising the longest stage and FDA review the shortest one.

5.2. Time from Phase Start to Next Phase Start

The start-to-start parameter refers to the elapsed time between the start of one development phase (e.g., Phase 2) and the start of the next development phase (e.g., Phase 3) supporting an application. For the non-clinical phase to Phase 1 estimate, we assumed that Phase 1 will begin immediately upon successful completion of the non-clinical development phase and receipt of IND approval from FDA. Similarly, for the FDA review to approval estimate, we used the estimates reported in Table 3 by therapeutic area (ranging from 9.6 months for oncology to 31.7 months for pain and anesthesia drugs). For the clinical phases, work on the clinical phases may overlap. In other words, the sponsor may begin Phase 2 clinical trials before completing the Phase 1 clinical trials. DiMasi et al. (2016) estimated the average phase duration, t_i, and average time to next phase, t_i−j, where i = 1, 2, 3, and j = 2, 3, FDA BLA/NDA review, for each of the three clinical phases as:

▪

t₁ = 33.1 months; t₁₋₂ = 19.8 months

▪

t₂ = 37.9 months; t₂₋₃ = 30.3 months

▪

t₃ = 45.1 months; t_{3−FDA BLA/NDA Review} = 30.7 months

To estimate the average phase start to next phase start durations, we used the DiMasi et al. (2016) estimates along with our phase duration estimates. For example, the average Phase 1 length for the Anti-Infective therapeutic area is 21.5 months (Table 3). Then, we estimated the average time to Phase 2 as the product of estimated average Phase 1 length (21.5 months) and the ratio of average time to Phase 2 to average Phase 1 length (19.8 ÷ 33.1 months) as reported in DiMasi et al. (2016) at 12.9 months (= 21.5 × [19.8 ÷ 33.1] months).

Table 3

Average Phase Durations (in Months), by Therapeutic Area.

As can be observed from Table 2, there is overlap between successive stages of clinical development. For example, sponsors begin Phase 2 studies on a larger cohort of patients with more diverse conditions when initial safety and dosing results from Phase 1 studies are available even if those studies may not be fully complete. Thus, even though a Phase 1 study is estimated to last around 27.8 months on average across all therapeutic areas, a sponsor may begin a Phase 2 study on average 16.6 months after initiating the associated Phase 1 study.

5.3. Average Number of Patients Enrolled per Trial

Number of patients enrolled in a study is the largest single factor driving study costs (Moore, et al., 2020). We used three databases (Medidata, clinicaltrials.gov, and FDA DASH), of which FDA DASH and Medidata are non-public, to estimate the average number of patients enrolled per trial by therapeutic area and phase (Table 4). The databases used cover different periods and vary in sample size, i.e., number of studies included. Ideally, the average number of patients enrolled estimate should be based on recent trials (preferably in the last 5 years) conducted in support of an NDA or BLA submission to FDA and rely on a large number of trials for each therapeutic area. None of the three databases satisfy these criteria fully. For example, Medidata database includes large number of studies, but it covers studies from 2004 through 2012 and includes trials that are not conducted in support of an NDA or BLA application to FDA. Similar to Medidata, clinicaltrials.gov database has a large number of studies from 2014 through June 2020 but also includes those that are not conducted in support of an NDA or BLA. On the other hand, FDA DASH database includes information from more recent trials (2007 through 2017) that are conducted in support of an FDA application but has fewer studies⁸ and does not include data on Phase 1 or Phase 4 trials or those trials that failed. Thus, we used all three databases to calculate the weighted average number of patients enrolled by therapeutic area and phase where the weights are the number of studies in each database.

Given the proprietary nature of information used, Table 4 only depicts the weighted mean number of patients per trial by therapeutic area estimated, where the weights are the number of studies in each data source relative to the total number of studies across all sources.

From Table 4, the weighted average number of patients per trial across different therapeutic areas are highly variable. For Phase 1, the weighted average ranges from 31 patients for hematology to 121 for ophthalmology trials; 133 for dermatology to 323 immunomodulation trials for Phase 2; 233 for hematology to 1,209 for pain and anesthesia trials for Phase 3; and 261 for oncology to 1,430 for anti-infective trials for Phase 4. Across all therapeutic areas, the weighted average number of patients enrolled per trial is 51 for Phase 1, 235 for Phase 2, 630 for Phase 3, and 708 for Phase 4.

Table 4

Average Number of Patients per Trial, by Therapeutic Area.

Further, within several therapeutic area and phase combinations, the variation across the average number of patients reported in the different databases is also significant. For example, the average number of patients in Phase 3 cardiovascular trials in FDA DASH is over nine times larger than that estimated from clinicaltrials.gov and over five times larger than that estimated from Medidata. However, there are a few therapeutic area and phase combinations for which this variation is minimal, such as Phase 2 and Phase 3 dermatology trials.

5.4. Average Number of Trials Conducted in Support of an FDA NDA/BLA Application

Sponsors indicate whether a trial is associated with an IND when they register it in clinicaltrials.gov. However, this information is only available to the National Institutes of Health (NIH) and the FDA, not to the general public. Thus, we requested a custom data pull from FDA CDER to estimate the average number of trials per IND application. FDA’s internal tracking system allows drug application reviewers to select from over 800 IND Division Class Codes (Tier 3), which are mapped onto 43 broader (Tier 1) division class categories. We mapped our therapeutic areas to these 43 FDA categories and FDA CDER compiled the number of INDs and IND-linked clinical trials by these therapeutic areas and phase. Next, FDA CDER calculated the average number of trials by therapeutic area and phase by dividing the clinical trial counts for a given phase and therapeutic area by the unique IND counts for the same phase and therapeutic area. FDA CDER’s (2019d) estimates are provided in Table 5.

From Table 5, the average number of trials conducted in support of an FDA application for a new drug is 1.71 for Phase 1, 1.52 for Phase 2, 2.66 for Phase 3, and 1.64 for Phase 4 across all therapeutic areas. For most therapeutic areas, sponsors conduct more than the two required Phase 3 trials with some running over four (endocrine) Phase 3 trials.

Table 5

Average Number of Trials Conducted in Support of an FDA NDA/BLA Application, by Therapeutic Area and Phase.

5.5. Average Cost per Patient

The total cost of a clinical trial for a given phase and therapeutic area, C_total, includes study-level costs (such as institutional review board approvals and source data verification costs), C_study, patient-level costs (such as recruitment and clinical procedure costs), C_patient, and site-level costs (such as monitoring and project management), C_site (Sertkaya, et al., 2016), i.e.:

We used three different data sources to estimate the average cost per patient. Two of the data sources (Cutting Edge and Medidata) included data on total clinical trial costs and the number of patients enrolled which allowed us to directly estimate the average cost per patient using the above equation. The third source, IQVIA, only contained information on patient-level costs, which comprise between 10 to 70 percent of total trial costs depending on therapeutic area and phase according to information available from the Medidata database. For comparability, we adjusted the reported IQVIA patient-level costs by these percentages. For example, if IQVIA reported a patient level cost of $10,000 for a Phase 1 study and patient-level costs were estimated to be around 20 percent of total costs in Medidata for that therapeutic area, we estimated the IQVIA average cost per patient at $50,000 (= $10,000 ÷ 0.20). The approach assumes that the shares of study, patient, and site costs for IQVIA are equivalent to those in Medidata. Due to the proprietary nature of these databases, we only present the weighted average cost per patient estimates by therapeutic area and phase in Table 6, where the weights are the number of studies in each database. As expected, the average cost per patient varies significantly by therapeutic area; $19,399 (anti-infective) to $349,363 (hematology) for Phase 1, $41,323 (cardiovascular) to $100,554 (hematology) for Phase 2, $30,001 (anti-infective) to $118,473 (hematology) for Phase 3, and $13,814 (anti-infective) to $56,824 (endocrine) for Phase 4. Across all therapeutic areas, the average cost per patient is $81,338 for Phase 1, $58,618 for Phase 2, $53,180 for Phase 3, and $35,190 for Phase 4 trials.

Table 6

Average Per-patient Costs (in 2018 $), by Therapeutic Area and Phase.

5.6. Phase Transition Success Probabilities

The phase transition success probability parameter represents the probability of a sponsor successfully moving from one stage of drug development to the next. If, for example, out of 100 new drug candidates that make it to Phase 1, 30 successfully proceed to Phase 2, then the phase transition probability from Phase 1 to Phase 2 is 30 percent. We used published studies to estimate the average phase transition success probabilities (Table 7). Across all therapeutic areas, successfully transitioning from Phase 2 to Phase 3 generally has the lowest likelihood at 35.9 percent (ranging from 27.4 percent for respiratory system to 56.6 percent for hematology). Getting approval from the FDA for a new drug that has cleared Phase 3 has on average 88.3 percent likelihood across all therapeutic areas. Further, only 8.5 percent (= 0.68 × 0.602 × 0.359 × 0.655 × 0.883) of new drug candidates successfully move from non-clinical development to market. However, as the drug candidate successfully clears each successive development stage, the odds of making it to market improve. As expected, there is variation in this likelihood across therapeutic areas with hematology drugs having the highest likelihood at 17.8 percent and oncology drugs having the lowest likelihood at 4.1 percent (not shown).

Table 7

Transition Success Probabilities, by Therapeutic Area and Phase.

5.7. Opportunity Cost of Capital

The opportunity cost of capital (OCOC) represents the rate of return (net of inflation) that the sponsor would otherwise be able to earn at the same risk level as the investment in the new drug that has been selected. Some critics have argued that “innovative companies must do R&D, and this is a regular cost of doing business; so estimated profits foregone should not be added to out-of-pocket costs. If revenues are coming in from other products, then the [R&D] costs are recovered as one goes along” (Light & Warburton, 2011). Others have questioned whether the appropriate cost of capital should be as high as 11 percent, the value used in several studies from the Tufts Center for the Study of Drug Development (Tufts CSDD).

As described by Chit, et al. (2015), there is an opportunity cost associated with the use of capital, which is a scarce resource, and this cost needs to be accounted for in estimating drug development costs. The value of OCOC can vary significantly by sponsor-specific factors, such as product portfolio, venture capital funding, and size of company, as well as other exogenous factors, such as economic and regulatory climate for drug development projects. There are accepted methods in finance for estimating the opportunity cost of capital for different economic sectors and firms, including the capital asset pricing model (CAPM), and the Fama and French (F-F) 3-factor model. The CAPM model is the most widely used approach (Chit, et al., 2015).

There are several CAPM studies that evaluated OCOC for the biopharmaceutical market as a whole as well as some broad sub-sectors, such as small and large molecules. Table 8 presents the different OCOC estimates available from the published literature. For the model, we used 11 percent as the OCOC for drug development projects, which is the average of figures reported for the biopharmaceutical industry as a whole.

Table 8

Published Estimates of Opportunity Cost of Capital.

5.8. Out-of-Pocket Cost Estimates

We calculated the total out-of-pocket cost by phase and therapeutic area as the product of per-patient cost (CPP), average number of patients enrolled per trial, and the average number of trials. The out-of-pocket cost for the FDA BLA/NDA review and approval was estimated at $2.6 million, the published FDA fee for an application requiring clinical data for fiscal year 2019 that spans October 1, 2018 through September 30, 2019 (Federal Register, 2018).

To estimate non-clinical cost, we adopted the approach by DiMasi et al (2016). There are no published data on non-clinical costs per drug candidate. Pharmaceutical companies have long claimed that it is difficult to attribute non-clinical R&D expenses to drug candidate compounds. In their 2016 study, DiMasi et al. estimated the ratio, R, of preclinical to clinical expenditures based on aggregated data on preclinical spending and assumptions around the duration of preclinical testing. Based on the reported amounts in Figure 2 of that study, they estimated the preclinical and clinical costs at $430 million and $965 million in 2013 dollars per approved drug, which translates to a ratio of 44.6 percent (DiMasi, et al., 2016). These estimates were based on data voluntarily submitted by anonymous biopharmaceutical companies as well as proprietary databases. The specifics of how they calculated this ratio is neither fully detailed in their study nor is available in other studies that are in the public domain. Thus, similar to other studies on this topic, we relied on the same reported ratio, 44.6 percent, to estimate non-clinical out of pocket costs per approved drug, which were then translated to a cost per drug candidate basis using the estimated aggregate mean success to approval rates by phase. More specifically, given that the estimated Phase 1, 2, and 3 costs are C₁, C₂, and C₃) and the estimated probability of approval from a given phase, i, is P_i, then the expected non-clinical stage cost, E(C_{non−clinical}), per approved drug was calculated from equation (1) as:

Then, using equations (1) and (11), the non-clinical cost per drug candidate was calculated as:

Given the sizable impact of non-clinical cost on overall cost of drug development, we also conducted a sensitivity analysis by varying this value +/-10 percent (Table 9). As can be observed from the table, the change in this ratio results in a proportionate change in expected capitalized cost estimate but a less than proportionate change in mean out-of-pocket cost estimate.

Table 9

Sensitivity Analysis of Varying Assumptions on Non-clinical to Clinical R&D Ratio, R, on Overall Cost Estimates.

5.9. Results

5.10. Baseline Development Cost Estimates

According to published studies that rely on proprietary data, the cost of drug development could range from $314 million to $2.8 billion (in 2018 dollars) depending on the therapeutic area, the cost of capital or phase transition success rate assumptions used in the modeling (DiMasi, et al., 2003; Jayasundara, et al., 2019; Mestre-Ferrandiz, et al., 2012; Adams & Brantner, 2006; Adams & Brantner, 2010; DiMasi & Grabowski, 2007; DiMasi, et al., 2004; DiMasi, et al., 2016; Paul, et al., 2010). Recent studies that have used publicly available data (mainly data reported by biopharmaceutical companies to the Securities and Exchange Commission in their annual 10-K and Quarterly 10-Q filings) report cost figures that range from $734 million for cancer (Prasad & Mailankody, 2017) to $4,461.2 million for antineoplastic and immunomodulating agents (Wouters, et al., 2020) (see Table 10).

Our analysis suggests that the average out-of-pocket cost of developing a new drug is around $131.8 million before conducting post-approval studies, and approximately $172.7 million when post-approval studies are accounted for (see Table 11). Of those costs inclusive of post-approval Phase 4 studies, 7 percent is non-clinical stage related, 68 percent is clinical stage (i.e., Phase 1, 2, and 3) related, 2 percent is review phase, and the remaining 24 percent is associated with post-approval stage, which includes Phase 3 follow-up studies, where applicable, and Phase 4 post-marketing studies. When capitalized to account for cost of capital and after accounting for the costs of failures, expected capitalized average development cost for new drug development is approximately $844.6 million before conducting post-approval studies and $879.3 million after conducting them. These development costs vary widely depending on therapeutic area as shown in Table 11. At one end of the spectrum are anti-infective drugs that cost about a third of this estimate ($378.7 million including post approval study costs) and at the other end are pain and anesthesia drugs that are more than four times as costly to develop ($1,756.2 million including Phase 4 costs).

Table 10

Recent Published Estimates of Cost of Drug Development.

Table 11

Average Cost of Developing a Drug for the U.S. Market (in Million $ 2018).

As indicated, expected capitalized costs are higher than out-of-pocket costs because they consider the opportunity cost of capital that embodies the time value of money and the fact that there will be failures along the way. The figures presented in Table 11 represent our baseline cost of new drug development against which we evaluate different strategies designed to improve likelihood of phase transition success and/or reduce non-clinical, clinical, FDA NDA/BLA phase, and post-approval related costs and durations.

As Table 11 illustrates, the primary driver of development cost is clinical stage followed by non-clinical stage expenditures when we account for cost of failures and cost of capital. From a capitalized out-of-pocket cost perspective that takes account of the time value of the investment but not failure costs, non-clinical and clinical development stages account for 13 percent and 70 percent of total capitalized development costs, respectively, whether or not post-approval Phase 4 study costs are included.

From an expected capitalized cost perspective in which both cost of failures and the time value of the investment are incorporated, the share of total expected development cost represented by the non-clinical stage is 40 percent, inclusive of post-approval study costs. Non-clinical stage represents the second largest portion of total expected capitalized development costs following the clinical stage at 53 percent primarily because the probability of moving from non-clinical stage to a marketable drug is only 8.5 percent on average. Thus, the $11.8 million and nearly 3 years needed to conduct non-clinical studies are much greater in real economic impact than their nominal value suggests. As the drug developer successfully transitions from one development stage to another, the likelihood of approval hence expected returns change. Even though a large, Phase 3 study may be more expensive out-of-pocket than non-clinical work, the odds of a drug candidate making it to market is significantly higher (65.5 percent) if the new drug candidate has already cleared the non-clinical, Phase 1, and Phase 2 stages than one that is at the target identification stage (8.5 percent).

The clinical phases of drug development (Phase 1, 2, and 3) are the largest contributor to total out of pocket development costs, comprising around 68 percent of total costs inclusive of post-approval studies. From a capitalized out-of-pocket cost perspective, clinical development comprises 70 percent of total capitalized development costs, including post-approval costs but excluding the time value of the investment. From an expected capitalized out-of-pocket cost perspective, the share of total expected capitalized development costs represented by clinical development is around 53 percent, including post-approval study costs. Phase 3 costs constitute the vast majority of clinical development costs, due primarily to enrolling large number of patients (approximately 630 versus 51 for Phase 1), taking longer than Phase 1 (38.0 months versus 27.8 months), and greater out-of-pocket costs (approximately $89.3 million vs. $7.1 million).

It is important to note that the estimated costs presented in this study do not include some significant elements, such as development of chemistry, manufacturing, and controls (CMC), and manufacturing plant design and build, which could be significant.

5.11. Impact of Select Clinical Trial Strategies on the Total Cost of Drug Development

As described in our previous study (Eastern Research Group, Inc., 2022), we asked our panel of experts to evaluate the impact of various clinical study strategies on the cost, duration, and phase transition success probability of drug development stages. A summary of our experts’ estimates is presented in Table 12 and estimates by therapeutic area are provided in Appendix A. Negative percentages indicate reductions in a given parameter (e.g., use of mobile technologies would reduce clinical study costs, on average, by 3 percent during Phase 1 holding all other factors constant), and positive percentages indicate increases in a given parameter (e.g., using biomarkers as surrogate endpoints would increase a developer’s probability of successfully transitioning from Phase 2 to Phase 3, on average, by 2 percent across all therapeutic areas holding all other factors constant).

We then evaluated the overall impact of each strategy on total expected development cost (see Table 13). Using our total expected capitalized cost (including post-approval studies) estimates as our baseline, we evaluated the change (or delta [Δ]) to this total expected cost if a developer were to implement a given strategy across all therapeutic areas. For each strategy, we evaluated the reduction in overall expected development cost attributable to the cost savings, time savings, and increases in phase transition success probability associated with that strategy. For example, use of adaptive design in clinical trial protocols are associated with sponsor overall cost savings of 22.8 percent, time savings of 1.6 percent, and a phase transition success probability increase of 19.2 percent (Table 13). When incorporated into our drug development cost model, these changes result in a total expected capitalized development cost of $678.7 million, which is 22.8 percent lower than our baseline estimate of $879.3 million.

From Table 13, the strategy with the largest impact on overall development costs across all therapeutic areas is Improvements in FDA Review Process Efficiency and Interactions (−27.1 percent), followed by Adaptive Design (−22.8 percent), and implementation of a Simplified Clinical Trial Protocols and Reduced Amendments (−22.2 percent). Those strategies with the lowest expected development cost savings across all therapeutic areas include Use of Patient Registries (−9.9 percent), Biomarkers as Surrogate Endpoints (−13.3 percent), and Electronic Health Records (−13.6 percent).

Table 12

Expert Estimates of Strategy Impacts on Cost, Duration, and Probability of Phase Transition Success for Drugs (All Therapeutic Areas Combined).

Table 13

Impacts of Clinical Trial Strategies on Baseline Cost, Duration, and Phase Transition Success Probability – Drugs.