An evaluation of the US public–private Combating Antibiotic-Resistant Bacteria Accelerator (CARB-X) initiative

Abstract

Introduction

Developing novel antibiotics to replace ineffective existing ones is a priority policy objective. The Combating Antibiotic-Resistant Bacteria Accelerator (CARB-X) is the largest global public–private initiative to advance this goal. The US-based CARB-X has committed over $1 billion US dollars through 2028 to innovators, and provides financial and technical support for the development of novel antibiotics through a competitive request for proposal application process.

Methods

We evaluated the outcomes of all applications to CARB-X between the program's inception in 2016 through 2019. Outcomes were evaluated through February 2024 and we used a quasi-experimental design controlling for application quality using multivariate logit regression.

Results

We found that 77 applications were funded and 59 applications were not funded. CARB-X–funded projects, controlling for funding score, were more likely to advance in the clinical development phase, but other outcomes were positively but not statistically significantly associated with CARB-X support, compared with unfunded projects. No applications had advanced to regulatory approval or commercial launch.

Conclusion

Our results suggest that CARB-X is positively associated with some intermediate project outcomes. Future work should examine later-stage outcomes associated with CARB-X. Evidence-based initiatives are needed to address the threat of antibiotic resistance.

Introduction

The effectiveness of many antibiotics is rapidly diminishing, as bacteria become increasingly resistant due to natural evolution and decades of extensive antibiotic use.^1,2 The effects of antibiotic resistance (more generally, antimicrobial resistance [AMR]) are devastating; AMR is a leading cause of morbidity and mortality, with the highest burdens in low-resource settings.^3-6 There were an estimated 4.95 million global deaths associated with AMR in 2019, of which approximately one-quarter are directly attributed to AMR.^7,8 Diminishing antibiotic effectiveness is leading to the resurgence of existing communicable disease threats and the emergence of novel ones, as illustrated by the increasing prevalence of hard-to-treat multidrug infections.^8,9 Projections suggest an escalation in AMR-associated deaths without additional intervention.⁹

Effectively combating AMR requires a multifaceted approach that thwarts the spread of infectious disease, facilitates sustainable and equitable use of antibiotics when indicated, and preserves existing therapies.^8-13 Developing novel antibiotics to replace ineffective existing ones, or to target “priority pathogens” from the World Health Organization (WHO) and the US Centers for Disease Control and Prevention (CDC) lists for which lack of treatment may result in severe adverse consequences, is a critical component of the solution, including but not limited to antibiotics to address gram-negative infections.^8,9,14

Nevertheless, the advent of new antibiotics has been slow and the current clinical pipeline is not robust enough to address the clinical need.^15-19 Few antibiotics that provide a significant clinical benefit over existing drugs have been approved in the last 30 years.^17,18 Between 2017 and 2021, 12 new antibiotics were approved by the US regulatory authority (the US Food and Drug Administration [FDA]) or the European authority (the European Medicines Authority [EMA]), with only 7 aimed at a priority pathogen.¹⁹ Additionally, there are less than 3 dozen antibiotics in clinical development targeting WHO priority pathogens.¹⁹ However, many of these products lack any innovative chemical structure, or confirmed activity against priority pathogens.¹⁹ Among projects in the preclinical antibacterial pipeline, most are affiliated with privately owned and pre-revenue companies,¹⁵ which are at higher risk of failure compared with publicly traded established firms.

While scientific barriers are substantial, economic barriers may present the biggest hurdle to novel antibiotic development.^18,20 Good stewardship requires that a novel antibiotic agent be held in reserve to inhibit the development of resistance to the new agent and for use against the most difficult cases, thus keeping the volume of use and sales low. Furthermore, hospital-based antibiotics are generally reimbursed to the provider by the payer under a Diagnosis-Related Group (DRG) bundled payment, providing an incentive for providers to prefer inexpensive generics.²¹ Finally, antibiotic infections generally occur over a short period of time, after which the patient is no longer in need of the drug, contrary to chronic diseases that tend to generate revenue over the span of the patient's life postdiagnosis.^20,22 Consequently, expected sales quantity and associated profits of novel antibiotics to treat priority pathogens are low.

In the past decade, governments and private foundations have created initiatives to address these concerns.^15,23-25 The Combating Antibiotic-Resistant Bacteria Accelerator (CARB-X) is the largest, currently operating global initiative to address AMR.²⁶ Co-founded by a US government agency, the Center for the Biomedical Advanced Research and Development Authority (BARDA), and an international philanthropic organization, the Wellcome Trust, in 2016, CARB-X is funded by additional national governments and charitable foundations. Since its inception through 2028 CARB-X has total committed funding exceeding $1 billion US dollars (USD) to advance the mission of accelerating a diverse portfolio of high-quality antibiotics towards clinical development. CARB-X provides financing and technical assistance, including business support, mentoring, and research support services, to developers for defined projects aimed at expanding the AMR armamentarium, including therapeutics, diagnostics, and preventatives for bacterial infections. CARB-X supports projects in the early stages of development, from the Hit-to-Lead, Lead Optimization, and Preclinical research stages up to demonstration of safety in humans in phase 1 clinical trials. CARB-X grant recipients may include university research centers, small private companies, and large publicly traded companies. Since its inception through January 2024, CARB-X had spent $317 million USD supporting early-stage antibiotic therapeutic and prevention programs, representing a large proportion of concurrent global funding for preclinical AMR-targeted product development.^15,26

Understanding the impact of CARB-X and other initiatives encouraging novel antibiotic development is important. A study by Viera, Kimmet and Moon (2021) compared the costs of development associated with initiatives aimed at encouraging the development of new antibiotics, but did not provide an evaluation of existing efforts.²⁷

We used quasi-experimental methods to evaluate the impact of receiving financial and project assistance support from the CARB-X initiative, comparing outcomes of funded and unfunded projects and controlling for project quality based on an objectively assigned score of each project application. Since CARB-X is a new initiative and drug development is long and risky—the mean time from Hit-to-Lead to first drug approval for antibiotics is 12.5 years²⁸—we focus on early-stage intermediate development outcomes measured at the application level.

Data and methods

Overview of analytical approach

This was an independant evaluation of the effect of CARB-X support on intermediate project outcomes. The hypothesis we tested was that CARB-X financial and technical support did not have an effect on funded project outcomes. CARB-X leadership provided the study team with data for the evaluation, but CARB-X leadership and staff were not involved in study design nor result interpretation.

The primary unit of our evaluation was the application, and we first separated applications into those that were awarded funding and those that were not awarded funding. We compared characteristics between funded and unfunded applications, finding evidence that they differed substantially. Moreover, we recognized that the quality of the application might be strongly associated with funding, which would confound our assessment of the effect of CARB-X support on outcomes. To address this concern, we followed other previously published studies that have examined the effect of receiving grant support on project outcomes by controlling for project quality based on an objectively assigned score of each funded and unfunded project application.^29,30 For example, a recent article studied the effects of the US Department of Energy's small-business program using a similar quasi-experimental approach, leveraging the availability of application scores, and reported that early-stage funding support positively affects patents and firm revenue.³¹ The process for applicant selection into the CARB-X program is competitive and involves a formal request for proposals (RFP) mechanism. Project developers submit an application, which is evaluated and scored by an external advisory board composed of experts in infectious disease, antibiotic drug development, and commercialization, on the strengths of the scientific promise, conformity of product characteristics with funding priorities, and assessed ability of the team to execute on predetermined milestones. We narrowed our focus to examining the early-stage outcomes of funded vs unfunded applications that had passed a first application screen and qualified for evaluation in a second (“long-form”) screen. We then examined the correlation between application funding outcome and funding score, finding evidence that funding score is strongly and positively correlated with funding but not perfectly determinative of funding and varied with other application characteristics (See Table S3). We believe this is for 3 reasons (see Table S3). First, the advisory board assigning funding scores to project applicants was separate from the governing board making offers of funding to applicants (see later for details). Second, the CARB-X budget is limited and each RFP had unique specifications and priorities. Consequently, projects with the same or similar scores assigned by the advisory board could receive different funding offers by the governing board with these considerations in mind (see later for details). Third, the highest-quality projects likely had outside support opportunities, while the marginal projects were more reliant on CARB-X funding decisions. The CARB-X governing board took the availability of alternative funding into account when advancing offers of funding to applicants, while the advisory board judged applications on their merits alone. As evidence of this effect, we observed that some very-high-scoring projects did not receive funding offers from CARB-X. Consequently, in our main analyses, we used multivariate logistic regression to estimate the effect of CARB-X funding on the binary outcomes controlling for funding score. We transformed the coefficients from the logistic regressions to odd ratios using the exponentiated method to facilitate interpretation. We interpreted the results on funding status as the marginal effect of CARB-X funding on outcomes after controlling for project quality. Since our analytic approach requires an assumption of linearity in the funding score, we used winsorization to limit the effect of very-high- and very-low-scored applications on outcomes (details below).

Data description and collection

Data on all project applications to CARB-X between 2016 and 2019, application scores, and ultimate funding awards for each application (including the disease, product, company, and timing of the award) were collected by CARB-X. Specifically, CARB-X collected and verified details about the product and company from the applicants via an electronic system that stored the data in a relational database since CARB-X began. Information about the timing of the funding round the applicants applied for, funding round RFP specifications (eg, restricted to products targeted to selected pathogens), application scores and thresholds for award recommendation and provision by the CARB-X evaluation committee, and progression of applications towards development was recorded in the CARB-X relational database. CARB-X staff used a variety of verification and quality checks to ensure data reliability and consistency in coding across years and multiple funding rounds and applications. These data were collected and updated as of February 12, 2024, and provided to the study team.

Additional details of the CARB-X application process and scores used in our analysis

We evaluated applications that were recommended for funding by the CARB-X executive board based on the results of their long-form application evaluation. Until 2019, applications to CARB-X awards entailed a short-form application and, if recommended for funding consideration, a subsequent long-form application. The short form was reviewed by a panel of external subject matter experts (SMEs). Their assessment of each application was based on a rubric with 3 categories: ability to address unmet medical need, strength of the science proposed, and capabilities of the team. The reviewers provided a score (excellent, good, fair, poor) in each of these 3 categories and an overall recommendation for funding (recommended or not recommended). After the SMEs submitted their individual scores, they convened for a discussion about the scores across the group. Following the discussion, the reviewers were allowed to change their scores and their overall recommendation. Applications that were approved at the short-form stage were then invited to submit a long-form application. Long-form applications were scored individually by a larger panel of external SMEs, referred to as a CARB-X advisory board. The scoring structure was the same as with the short-form application. The reviewers provided a score (excellent, good, fair, poor) in each of the 3 categories and an overall recommendation for funding (recommended or not recommended). The total number of reviewers per long-form application differed within the same RFP and between RFPs over time. Every project team that was invited to submit a long-form application was also invited to a presentation to the advisory board. The meeting with the advisory board includes time for an uninterrupted presentation, a question-and-answer period, and then a private discussion period within just the advisory board, without the presenters. After the presentation and discussion, advisory board members had an opportunity to change their scores and/or their overall recommendation. The recommendations from the CARB-X advisory board were presented to the governing board for CARB-X. The governing board evaluated the recommendations from the CARB-X advisory board in terms of strategic alignment with CARB-X objectives, especially considering other projects in the portfolio, budget available, and other considerations including funding from alternative sources. The CARB-X governing board made the final decision on application funding. Eleven long-form application variables and their definitions were provided to the research team and are described in Table S1.

We operationalized each long-form application score to represent the percentage of individual reviewers recommending funding. There were 6 funding score categories (0–5), as described in Table S2. Projects with 0% of individual reviewers recommending funding received a score of 0. Similarly, projects with 100% of individual reviewers recommending funding received a score of 5. Scores of 1–4 were given to projects with between 0% and 25%, 25% and 50%, 50% and 75%, and 75% and 100% of reviewers recommended funding, respectively.

Outcomes

We examined outcome measures that represent interim successes or advancement of the project or firm. We focused on early-stage outcomes, including successful transitions of the applications from 1 phase of development to another (eg, Hit-to-Lead to Lead Optimization), transition into human clinical trials, the continued operation of the firm and the project, and counts of filed and awarded patents. The first 2 of these outcomes are reported in the CARB-X database, while the remainder were obtained by the study team through other public sources (see later for details). We created a composite binary measure aggregating “any” or none of the single item outcomes. Then each of these outcomes was analyzed using a separate regression. For all outcomes of interest, the measure assessed the outcome from the time of funding award to February 2024 for funded applicants and from the time of application to February 2024 for nonfunded applicants.

It is possible that non–CARB-X sources of funding and support could contribute to project success, but we note that these outside influences might impact both CARB-X–funded and nonfunded projects. Applicants may receive outside funding and this may lead CARB-X to decline the application and the receipt of a CARB-X award may itself help attract additional investment capital into a project or firm. Unfortunately, we do not have access to all non–CARB-X sources of funding, and consequently we did not control for this potential independent effect on outcomes.

Firm data collection and outcome measurement

Firms were determined to be operational if there was evidence that the firm was still functioning, such as an updated copyright year on the company website or a recent press release on product status. In addition to company websites, LinkedIn, Crunchbase, and internet searches were also used to assess for evidence that a firm was still functioning. Similarly, firms that underwent mergers or were acquired were also included as operational, with the same sources mentioned previously being used to gather information. Firms were classified as small if they had 50 or fewer employees and “non-small” (ie, medium or large) if they had more than 50 employees.

Patent data collection and analysis

Patent outcomes for each application were obtained using the US Patent and Trademark Office (USPTO) Assignment dataset, which provides information on patent-asset conveyances recorded at USPTO during 1970–2021, including identifying information on assignee, dates of patent conveyance, identifying US patent and application numbers, and conveyance types.³² RStudio 4.3.1 was used to search for CARB-X applicant names as the assignee in the USPTO dataset. Patent information was exported to Microsoft Excel, where 1 of the authors MA (with supervision by the other authors) excluded patents not relating to antibacterial products. Since the USPTO dataset includes information on every time a patent application was updated as separate entries, remaining patents were checked using Google Patents to remove any duplicate patents or patent applications. Information on whether a patent application was filed, or a patent was granted, and the total number of patent applications filed, or patents granted, was then recorded.

Inclusion/exclusion criteria for analyses

Of the 136 projects that reached the long-form scoring stage, 17 projects were evaluated under a different scoring system (ie, group funding recommendation as opposed to individual reviewer recommendations) and, thus, these projects were excluded from analyses. Projects funded by CARB-X since funding round 8, including in its omnibus funding round in 2024, were excluded due to lack of meaningful time to evaluate outcomes of interest. Projects that were already in the clinical trial phase at the time of scoring were excluded from analyses evaluating advancement to clinical trials.

There were 4 projects and 9 total applications in which the same project reached the long-form scoring stage multiple times due to having applications in multiple rounds and, thus, these projects were included more than once in the CARB-X sample. Half of these projects had also been acquired by a different firm. To avoid double counting, these projects were only included in the outcome analysis on “whether the project remains active” once (only the latter and not the former iteration was included). However, these projects were included multiple times for other outcome analyses. For the outcome analyses assessing advancement in phase or to clinical trials, these 4 instances were treated as “separate” applications, with the timeline for the first application running from the time from first application to the decision point for the second application and the timeline for the second application running from the second decision point to the end of the data collection period.

Similarly, there were 28 unique firms that had multiple applications reach the long-form scoring stage. For the outcome analysis assessing whether a firm remained operational, these firms were included in the sample multiple times. Finally, because diagnostics represent such a materially different development path from therapeutics and preventatives, we restricted our analysis to therapeutics and preventatives.

Regression analysis of outcome data

Multivariate logit regressions for binary outcomes were run with an indicator for funded or not funded, controlling for funding score on outcomes for all applications that received a final recommendation for funding with the sample winsorized at 90%. Statistical tests performed were 1-sided and a P value > .05 was considered statistically significant. To perform these analyses, we used R package 4.31. The project was deemed human subjects exempt by Boston University's Institutional Review Board.

Results

Baseline characteristics

From the 1163 expressions of interest, 136 projects made it to the long-form scoring stage (the final evaluation stage). Table 1 describes the baseline characteristics of these projects. Eighty-eight percent of the projects targeted a gram-negative pathogen. One hundred percent targeted a pathogen on the CDC list, including 96% targeting an urgent or serious pathogen. Similarly, 93% targeted a pathogen on the WHO list, including 92% targeting critical or high pathogens.^31,33 More than 85% of projects were therapeutics. More than 90% were in one of the preclinical phases, including 37% in lead development. Almost 92% were from a private firm. There were 105 unique firms that had applications reaching the long-form scoring stage. More than 26% of unique firms had multiple applications reach the long-form scoring stage, with 3 firms having a maximum of 3 applications reaching this stage. More than 69% of firms were classified as small.

Table 1.

Baseline characteristics of CARB-X long-form projects and applicants.^a

	All long-form applicants (n = 136), No. (%)	Funded long-form applicants (n = 77), No. (%)	Nonfunded long-form applicants (n = 59), No. (%)
Targets a gram-negative pathogen?b	119 (87.5)	68 (88.3)	51 (86.4)
Targets a pathogen on CDC list?	136 (100)	77 (100)	59 (100)
Urgent	102 (75)	61 (79.2)	41 (69.5)
Serious	29 (21.3)	14 (18.2)	15 (25.4)
Concerning	5 (3.7)	2 (2.6)	3 (5.1)
Watch list	0 (0)	0 (0)	0 (0)
Targets a pathogen on WHO list?	126 (92.6)	72 (93.5)	54 (91.5)
Critical	108 (79.4)	62 (80.5)	46 (78.0)
High	17 (12.5)	10 (13.0)	7 (11.9)
Medium	1 (0.7)	0 (0)	1 (1.7)
Treatment modalityc
Therapeutic	116 (85.3)	68 (88.3)	48 (81.4)
Preventative	20 (14.7)	9 (11.7)	11 (18.6)
Funding roundd
1	31 (22.8)	18 (23.4)	13 (22.0)
2	21 (15.4)	14 (18.2)	7 (11.9)
3	14 (10.3)	10 (13.0)	4 (6.8)
4	37 (27.2)	16 (20.8)	21 (35.6)
5	15 (11.0)	9 (11.7)	6 (10.2)
6	13 (9.6)	6 (7.8)	7 (11.9)
8	5 (3.7)	4 (5.2)	1 (1.7)
Initial phase
Hit-to-Lead	37 (27.2)	20 (26.0)	17 (28.8)
Lead Development	49 (36.0)	34 (44.2)	15 (25.4)
Preclinical/IND-enabling	44 (32.4)	21 (27.3)	23 (39.0)
Phase 1 clinical Trial	6 (4.4)	2 (2.6)	4 (6.8)
Type of applicant
Private firm	125 (91.9)	73 (94.8)	52 (88.1)
Not-for profit	8 (5.9)	4 (5.2)	4 (6.8)
Public–private partnership	3 (2.2)	0 (0)	3 (5.1)
Unique applicant	105 (77.2)	63 (81.8)	54 (91.5)
Small firm (<50 employees)	94 (69.1)	54 (70.1)	40 (67.8)
Reviewer score (No. of applicants)e
Mean	64.6 (28)	81.8 (25)	41.6 (12)
Median	75 (28)	88.2 (25)	33.3 (12)
Mode	100 (28)	100 (25)	0 (12)
Range	100 (28)	90 (25)	100 (12)

Abbreviations: CARB-X, Combating Antibiotic-Resistant Bacteria Accelerator; CDC, Centers for Disease Control and Prevention; IND, Investigational New Drug; WHO, World Health Organization.

^aAll data on long-form projects and applicants were derived from each project's Expression of Interest application to CARB-X. The 2017 WHO and 2019 CDC priority pathogen lists were used for characterizing target pathogens.^31,32

^bTreatments with a broad-spectrum focus (ie, target both gram-positive and gram-negative pathogens) were included.

^cDiagnostics were excluded.

^dFunding round 7 was focused on only diagnostics and thus excluded.

^eRefers to percentage of individual reviewers recommending funding. Seventeen projects were judged using a group-based scoring system and did not have individual reviewer scores. These 17 projects—9 funded and 8 nonfunded—were excluded.

Assessment of funding outcomes by funding score

Of the 136 long-form applications, 77 were funded and 59 were not funded (Table 1). Of these 136 applications, 17 did not have individual reviewer scores. Of the 119 long-form applications with individual reviewer scores, the percentage that were funded varied from 16.7% for score = 1% to 89.3% for score = 5 (Table 2). Lower-scored applications were often funded when higher-scored projects were not, allowing us to examine outcomes as a function of funding without merely reflecting strength of score.

Table 2.

Funding outcomes by funding score among CARB-X applications of therapeutics and preventatives with a score (n = 119).^a

Funding score	Total No. of applications	No. of funded applications	Percentage of applications funded
0	12	0	0
1	12	2	16.7
2	13	5	38.5
3	26	15	57.7
4	28	21	75
5	28	25	89.3

Abbreviation: CARB-X, Combating Antibiotic-Resistant Bacteria Accelerator.

^aSeventeen total applications (including therapeutics and preventatives) were judged using a group-based scoring system and did not have individual reviewer scores. These 17 applications (9 funded and 8 nonfunded) were excluded.

Patent, project, and firm outcomes by funding status

Table 3 describes the outcomes of all CARB-X applications that reached the long-form scoring stage by funding status. CARB-X–funded long-form applications were more likely to have a patent application filed and patent granted, advance in phase, be from a firm that was still operational, and remain active compared with nonfunded long-form projects. Funded long-form applications also had a higher total number of patent applications filed and patents granted than nonfunded long-form applications. Of 77 funded applications, 39 advanced to the next stage of development (50.6%). Of 59 applications not funded, 14 advanced to the next stage of development (23.7%). Of 77 applications funded, 13 advanced into clinical trials (17.3%) and of 59 applications not funded, 9 advanced into clinical trials (16.4%). Of 77 funded applications, 428 relevant patents were filed (average of 5.6) and of 59 not funded, 212 relevant patents were filed (average of 3.6). Of 77 applications funded, 160 relevant patents were granted (average of 2.1) and of 59 applications not funded, 97 relevant patents were granted (average of 1.6). With regard to firm operations, of the 77 funded applications, 73 applicants remained operational at the time of analysis (94.8%) and of 59 applications not funded, 50 applicants remained operational at the time of analysis (84.7%). Two firms, which filed 3 applications in total for 2 funded projects, filed for bankruptcy; both projects were acquired and did not shut down. Similarly, of the 77 funded applications, 42 projects (55.3%) remained active while of the 59 not funded, 21 (37.5%) remained active.

Table 3.

Outcomes of CARB-X long-form projects and applicants.^a

	All long-form applicants (n = 136)	Funded long-form applicants (n = 77)	Nonfunded long-form applicants (n = 59)
Was a patent application ever filed?	92 (67.6)	55 (71.4)	37 (62.7)
Was a patent ever granted?	66 (48.5)	41 (53.2)	25 (42.4)
Total No. of patents filedb	640	428	212
Average No. of patents filedb	4.71	5.56	3.59
Total No. of patents grantedb	257	160	97
Average No. of patents grantedb	1.89	2.08	1.64
Did the project advance in phase?	53 (39.0)	39 (50.6)	14 (23.7)
Did the project reach clinical trials?c	22 (16.9)	13 (17.3)	9 (16.4)
Is the firm still operational?	123 (90.4)	73 (94.8)	50 (84.7)
Does the project remain active?d	63 (47.7)	42 (55.3)	21 (37.5)

^aValues are No. (%) unless otherwise indicated. Patent data were obtained using the US Patent and Trademark Assignment dataset and Google Patents. Data on advancement in phase were obtained using the CARB-X website, firm websites, and other relevant search results (eg, journal publications, press releases). Clinical trials data were obtained using ClinicalTrials.gov and the World Health Organization's International Clinical Trials Registry Platform. Data on whether a firm was operational were obtained using Crunchbase, Pitchbook, LinkedIn, firm websites, and other relevant search results (eg, news articles).

^bFirms with multiple projects were counted more than once.

^cSix applicants (2 funded and 4 nonfunded) that were in the clinical trials phase at the time of scoring were excluded.

^dFour projects (1 funded and 3 nonfunded) were projects that were included multiple times in the CARB-X sample due to acquisition of the project or applications in multiple funding rounds. These projects were excluded from this outcome analysis to avoid double-counting; n = 132 (76 funded and 56 nonfunded).

Regression analysis for patent, project, and firm outcomes

We ran separate multivariate logit regressions for the 7 binary outcome variables of whether a patent application was filed, whether a patent was granted, whether the application advanced in phase, whether the application advanced to clinical trials, whether the firm is still operational, whether the project remains active, and the 1 composite outcome variable holding constant funding score in the winsorized sample (Table 4). We found that the coefficient for the estimate of funding status was positive for the filing of patent applications, the granting of patent applications, advancement in phase, and the continuing operation of the firm, the continuation of the project, and the aggregate composite outcome variable. Only the advancement in phase outcome variable was statistically significant; our results suggest that CARB-X support is associated with 4.68 times the odds in the probability of advancement in clinical development compared with similar-quality unfunded projects. We also ran regressions with the full sample and the sample winsorized at 60%; results did not differ substantially and consequently we do not report the results of these additional analyses.

Table 4.

Multivariate logistic regression of outcomes on funding status for all CARB-X long-form projects and applicants with a score.^a

	Was a patent filed? (n = 119)	Was a patent granted? (n = 119)	Did the project advance in phase?b (n = 119)	Did the project reach clinical trials?c (n = 115)	Is the firm still operational?d (n = 119)	Does the project remain active?e (n = 115)	Composite outcome (n = 115)
Funding	1.073 [0.416, 2.768]	1.523 [0.614, 3.779]	4.682f [1.654, 13.257]	1.217 [0.356, 4.156]	1.702 [0.354, 8.182]	1.157 [0.465, 2.878]	6.439 [0, infinity]
Funding score	1.218 [0.911, 1.629]	1.088 [0.821, 1.443]	0.907 [0.664, 1.239]	0.967 [0.666, 1.404]	1.324 [0.845, 2.076]	1.165 [0.876, 1.550]	185.541 [0, infinity]

Abbreviation: CARB-X, Combating Antibiotic-Resistant Bacteria Accelerator.

^aValues are odds ratios [95% CI]. Diagnostics were excluded. Seventeen projects were judged using a group-based scoring system and did not have individual reviewer scores. These 17 projects (9 funded and 8 nonfunded) were also excluded.

^bProjects for whom it was unknown whether they advanced in phase were included as not advancing in phase.

^cFour applicants (2 funded and 2 nonfunded) that were in the clinical trials phase at the time of scoring were excluded.

^dFirms for which it was unknown whether they are still operational were included as not operational. Firms that underwent a merger or acquisition were included as operational.

^eFour projects (1 funded and 3 nonfunded) were projects that were included multiple times in the CARB-X sample due to acquisition of the project or applications in multiple funding rounds. These projects were only counted once in this outcome analysis to avoid double-counting.

^fSignificant at ɑ of 0.05.

Discussion

We used quasi-experimental methods to examine the marginal effects of receiving financing and technical support from CARB-X on antibiotic innovation outcomes. Our results suggest that CARB-X support was positively and statistically significantly associated with advancement in phase and positively but not statistically significantly associated with patent filing and award and continued operations of the firm compared with their unfunded counterparts who reached the long-form scoring stage. Our analytic approach ensures the greater success of funded projects can be attributed to CARB-X support and not to a set of inherently superior applications.

To our knowledge, this is the first effort to evaluate empirically the effect of new incentives to address the global AMR challenge. Consequently, it adds to an existing literature suggesting alternative incentives to support novel antibiotic development.^18,29,33,34 Our study also adds to the rather limited empirical economics literature on the effect of incentives for novel drug development generally.^35-44 Our work differs from previous studies that focus on the productivity of National Institutes of Health–supported or privately funded research and development investments generally.^38-43 Our work is more closely aligned with previous studies that have reported positive effects of incentives for innovation that target substantial unmet need, including the US Orphan Drug designation,³⁷ Pediatric Exclusivity,³⁸ Breakthrough designation,⁴⁴ and efforts to create medical countermeasures. Our study differs from previous evaluation efforts of these incentives in that our focus is on intermediate outcomes of drug development due to the limited time to evaluate CARB-X application outcomes. Pursuing a more robust analysis of drug development outcomes including approvals at a future time when the dataset will be richer by virtue of more observed applications and longer history of development is an important topic of future study.

The success of CARB-X helps address the AMR crisis by creating more promising drug candidates in the early part of the pipeline, an element that is needed.⁴⁵ As candidates move into phase II and phase III, they may find financing opportunities from downstream funding organizations such as BARDA, the newly established European Commission's Health Emergency Preparedness and Response Authority, the AMR Action Fund, or Global Antibiotic Research & Development Partnership. However, we still expect little appetite from established pharmaceutical companies or venture capital investors to fund these AMR product candidates through registration trials. This commitment may entail a hundred million dollars or more and consequently until and unless there is reason to believe that a commercial opportunity or “reward” will justify the investment. Thus, until funding to CARB-X is complemented by the establishment of a globally coordinated set of national or regional incentives, society remains at risk of not having the medicines it needs. Although many CARB-X–funded projects remain active, none has yet advanced to a market launch, suggesting that novel antibiotic development remains in a precarious state to address global priority needs.

Fortunately, there is progress in developing appropriate economic “pull” incentives to provide the necessary financial “reward” that CARB-X does not address. The United Kingdom leads, as its subscription pilot is transitioning into a permanent program, featuring a streamlined evaluation process to enhance accessibility and efficacy,⁴⁶ and Japan has initiated its own pilot for a national revenue guarantee system.⁴⁷ The Directorate-General for Health and Food Safety of the European Commission has proposed that EMA approvals of high-impact antibiotics should be rewarded with transferable exclusivity vouchers as part of its European pharma legislation revision.⁴⁸ This could be complemented (or supplanted) by a multicountry revenue guarantee system orchestrated by the Directorate-General for Health Emergency Response and Authority.⁴⁹ The Council of Canadian Academies has advocated for the adoption of a country-wide subscription model in a report commissioned by the Public Health Agency of Canada.⁵⁰ In recent years, the proposed US PASTEUR Act (Pioneering Antimicrobial Subscriptions To End Upsurging Resistance Act of 2023) has gained bipartisan sponsors, been introduced in multiple Congressional sessions, but has failed to pass and be implemented, while the Biden Administration included a subscription model akin to the one outlined in the draft bill in its 2024 budget.⁵¹ At the time of writing in Spring 2026, US government support for BARDA and the CARB-X initiative has been renewed and will continue until at least 2028, and CARB-X remains funded across many private and international sources as well.

Conclusion

To our knowledge, this is the first quasi-experimental evaluation of a contemporaneous initiative to develop novel antibiotics to address the global AMR threat. CARB-X provides financing and technical expertise to support novel antibiotic development, and is positively associated with project advancement in phase, and other intermediate outcomes, although they are not statistically significant. Future efforts should examine longer-term outcomes of initiatives such as CARB-X that generate early-stage antibiotic-development candidates, and the effectiveness of later-stage support programs and other existing product development partnerships to address AMR. Future initiatives might also aim to shore up resources to support CARB-X and related initiatives and complement such efforts with greater support for product commercialization.

Contribution statement

R.M.C. and E.J.E. conceptualized and designed the study, analyzed and interpreted the data, drafted the initial manuscript, reviewed and revised the manuscript for important intellectual content, and provided study supervision. R.M.C., E.J.E., and M.A. analyzed and interpreted the data, drafted the initial manuscript, and reviewed and revised the manuscript for important intellectual content. All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

Supplementary material

Supplementary material is available at Health Affairs Scholar online.

Data availability

The datasets used and/or analyzed for the study are available from the corresponding author on reasonable request.

Code availability

The underlying code for this study are available from the corresponding author on reasonable request.