The value of the accelerated approval pathway: real-world outcomes associated with reducing the time between innovations

Advances in pharmaceutical innovation are a leading contributor to overall life expectancy gains in the USA [1]. Innovation in the oncology therapeutic arena has been particularly rapid with the number of targeted therapy new molecular entities (NME) approximately doubling during 2018–2022 compared with 2015–2017 [2]. These advances in oncology drug therapy have led to improved survival for patients with cancer [3,4]. Between 2000 and 2016, new cancer medicines were associated with nearly 1.3 million deaths avoided for the 15 most common tumors [5].

In addition to the extended survival observed in the clinical trial setting, use of new cancer drugs in the real world may increase the chance that a patient remains alive to access future treatments still in clinical development. This concept of extending survival as a bridge to future, not yet approved, innovative therapies is also referred to as ‘real option value’ (ROV) and has been studied across different therapeutic areas. When the opportunity for patients to move from one life-extending treatment to a newly approved treatment is present, information on how this changes a patient's overall treatment trajectory and survival outcomes becomes increasingly important. Specifically, this information can impact how new treatments are valued by healthcare decision-makers and can influence treatment decision-making for oncologists [6–8].

Prior studies in oncology have found that this increased access to future innovations may provide an additional 5–20% survival to patients beyond the survival gains expected from their current life-extending therapy [9–12]. For example, in advanced melanoma, patients who initiated treatment with first-line ipilimumab and survived to access pembrolizumab via its accelerated approval (AA) gained 35% more survival (3.7 months) than patients in the ipilimumab clinical trial [13]. However, the magnitude of survival benefit may vary depending on the pace of clinical innovation, with a higher likelihood of this additional survival value seen in diseases with rapid clinical innovation where more patients are likely to access future innovations.

In the US market, one factor contributing to the pace of innovation has been the US FDA's AA program. The AA program is a regulatory pathway that allows the FDA more expeditiously to approve drugs that treat serious conditions with high unmet need based on a surrogate end point [14]. In recent years, the FDA has increasingly utilized the pathway to accelerate access to oncology products [15,16]. However, the AA pathway also has come under closer scrutiny given that not all medicines convert to traditional approval owing to withdrawals (i.e. olaratumab for soft tissue sarcoma), the cost of the medicines, as well as noted delays in completion of confirmatory trials for a handful of AA products [17–19]. Information on the net impact and value of the AA program in oncology is needed to better inform ongoing public discussion of the program's benefits and challenges [20]. Past research suggests that drugs approved through the AA program bring benefits based on modeled comparisons of health outcomes relative to standard of care, as well as analyses of aggregated clinical trial results [21,22]. However, there is no information on real-world outcomes for patients on life-extending therapies who access future drugs approved via the AA program.

This study sought to examine the role of the AA pathway in enabling ROV. Specifically, we isolate the impact of the AA pathway in enabling patients on life-extending treatments to access newly approved treatments as a next line of therapy. The objectives of this analysis were: (i) to estimate the number of patients receiving future treatments and their survival gain due to the AA pathway using a real-world database, and (ii) to extrapolate the results to the US population to estimate the overall impact of the AA pathway on total life-years (LYs). Given most of the recent AAs have occurred in oncology, we use real-world data for treatments in three different tumors as case studies to study the impact of this pathway on clinical outcomes.

Methods

Study design

In total, 36 previous AA indications were initially reviewed to assess their suitability as case studies for analysis, based on whether the AA drug was converted to a full approval and if the AA drug represented a potential follow-on line of care after a recent approval of a new drug in an earlier line of therapy. Case studies eligible for inclusion were further narrowed down based on the presence of an AA in the electronic health records (EHRs; cancer type had an AA between 2011 [date of data availability] and 2021), potential for ROV to be measured (cancer type had a newly approved ‘initial’ treatment in a previous line of therapy to the AA drug indication), data availability (cancer type available to the research team in the Flatiron Health database, availability of real-world progression-free survival data) and feasibility (sufficient sample size to conduct the analysis [i.e., rare tumors were excluded], sufficient follow-up time to estimate median progression-free and overall survival). Based on the selection criteria described above, metastatic melanoma, advanced bladder cancer and epidermal growth factor receptor positive (EGFR+) advanced non-small-cell lung cancer (aNSCLC) were considered eligible and selected as case studies. To assess real-world outcomes, we used the nationwide Flatiron Health EHR-derived longitudinal database, which is composed of de-identified patient-level structured and unstructured data, curated via technology-enabled abstraction [23,24]. We estimated patient-level outcomes for case studies in tumors with high levels of clinical innovation and for which recently approved products with extended survival could be eligible for future AA products within current survival estimates.

We estimated outcomes associated with the AA pathway by comparing observed outcomes to a counterfactual (hypothetical) scenario in which approval occurred at a later time through a traditional approval pathway (Figure 1). For our primary analysis, we assumed that the traditional approval would occur at the date of conversion from accelerated approval per the FDA-reported date of traditional approval. Two other counterfactual scenario analyses were conducted to understand how differences in the timing of full approval, including potential earlier off-label use after data presentations or a longer time to complete a confirmatory trial, impacted findings. For these scenarios, the date of approval was either (i) the average time to full approval conversion in oncology overall (i.e., a longer time to approval than the actual dates for our case studies) or (ii) the date of the initial data presentation (i.e., phase III study) at a clinical conference (i.e., a shorter time to approval than the actual dates for our case studies).

To assess the proportions of patients receiving future AA drug(s) and survival outcomes we followed up patients starting treatment with a newly approved drug (referred to hereafter as the ‘initial drug’) in the three selected tumor types during the pre-AA period, defined as the period after their approval until the date of the AA of the future drug (Table 1). The initial drugs evaluated included first-line ipilimumab (metastatic melanoma), post-platinum cancer immunotherapies (advanced bladder cancer), and first- and second-generation EGFR tyrosine kinase inhibitors (TKIs) (EGFR+ aNSCLC; Table 1). Patient inclusion criteria included treatment within 120 days of their advanced diagnosis, use of the initial drug(s) during the pre AA period, prior platinum therapy (bladder cancer only) and an EGFR+ test result prior to treatment (EGFR+ aNSCLC only; Supplementary Table 1). Future AA drugs studied in the subsequent line of treatment included cancer immunotherapies (metastatic melanoma), enfortumab vedotin-ejfv (advanced bladder cancer) and osimertinib (EGFR+ aNSCLC). The AA dates for each of the above drugs, as well as the assumed counterfactual scenario dates, can be found in Table 1. For the cases where there were multiple drugs, the date for the drug with the earliest approval date was used as the start of the pre-AA period or start of the AA period.

Outcomes

Patient-level outcomes

Outcomes measured from the real-world database included the proportion of patients: receiving the initial drug (by year of treatment initiation); surviving to the AA date for the future drug; and receiving either the future AA drug (treatment cohort A), standard of care (SoC) (treatment cohort B) or no subsequent treatment (treatment cohort C). For each of these cohorts, the mean overall survival (OS) was estimated. Given the limited sample size in metastatic melanoma, to ensure more stable survival estimates, we followed the methodology in a previous ROV analysis that expanded the patient pool for the survival analyses to include patients treated up until the date first-line pembrolizumab data were publicly presented [13], which may have initiated a change in SoC. For the treatment cohorts in which the initial drug was followed by the future AA drug or no subsequent treatment (cohorts A and C), mean OS was estimated from the time of initial drug initiation until the date of death. To estimate the OS for patients who received SoC after the initial drug, we constructed a survival curve by adding: the length of time from the start of the initial drug to the future AA drug initiation for each patient from cohort A; with mean time from SoC initiation to death from cohort B. This was done so that the OS differences between those receiving the future AA drug and SoC were not driven by differences in the length of time spent on the initial drug, but rather differences in OS from the future AA drug and SoC.

As we sought to understand how the timing of future AA availability impacts outcomes for patients on treatments before the AA date, we estimated the average survival gain among all patients on the initial drug to understand the overall impact of the AA pathway. Specifically, this average estimate reflects the probability of remaining alive to receive the future AA drug as well as the additional survival provided by the initial AA drug.

Statistical analysis

Baseline patient characteristics for the three cohorts in each tumor were summarized using descriptive statistics and compared using χ² test for categorical variables and the Kruskal–Wallis rank sum test for continuous variables. We generated Kaplan–Meier curves to estimate restricted mean OS for each cohort (A–C) in each tumor. Survival curves were also generated for patients within cohort B by line of therapy for which SoC therapy was initiated (for use in construction of survival curves for patients receiving SoC as described above). To mitigate potential confounding in comparison of survival estimates across cohorts, we weighted patients across cohorts using stabilized inverse probability of treatment weightings (IPTW). Weightings were constructed for propensity to use the future AA drug (cohort A) in relation to SoC (cohort B; for cohort A vs B comparison) and also for propensity to use the future AA drug (cohort A) in relation to no treatment (cohort C; for cohort A vs C comparison). Variables underlying the propensity calculations were age, sex, race and ethnicity, region, practice type (academic or community), payer type (available for bladder cancer and EGFR+ aNSCLC only), Eastern Cooperative Oncology Group (ECOG) performance status (available for EGFR+ aNSCLC and bladder cancer only), BRAF status (metastatic melanoma only) and type of cancer immunotherapy (CIT) used (pembrolizumab, atezolizumab or nivolumab/avelumab/durvalumab; for bladder cancer only). Payer type and ECOG performance status were not included in constructing IPTW for metastatic melanoma owing to a high level of missingness in the cohort. Statistical analyses and calculation of proportions were performed using R Statistical Software (v4.1.3; R Core Team 2022).

Population-level outcomes

Outcomes from the real-world data analysis were extrapolated to the population level by identifying the number of patients starting the initial drug in each year according to real-world observations of US population level estimates of drug utilization (Supplementary Table 2) [25]. The number of patients starting the initial drug at the population level was then multiplied by the proportion surviving to the AA date (estimated from the real-world database as described above) and categorized according to receipt of subsequent therapy to estimate the number and percentage of patients receiving the AA drug at the US population level. Outcomes attributable to the AA pathway were measured by estimating the change in the number and percentage of patients able to use the AA drug based on delays in access per the three counterfactual scenarios (primary analysis and two alternative scenarios) described above. The total LY impact at the population level was estimated by multiplying the additional number of patients by the average survival gain due to AA among all initial drug patients.

Results

Patient characteristics

In total, 353 patients with metastatic melanoma, 888 patients with advanced bladder cancer and 734 patients with EGFR+ aNSCLC were included in the analysis (Table 2 & Supplementary Table 1). Patient characteristics by tumor and treatment received are shown in Supplementary Table 2. Baseline patient characteristics, including age, gender, race and ethnicity and region, were similar between cohorts within individual tumor types. Payer type was primarily Medicare, although fewer Medicare beneficiaries received future AA drugs in advanced bladder cancer and EGFR+ aNSCLC.

Receipt of AA drugs in the real-world database

The proportion of patients on the initial drug surviving to the FDA AA date of the future drug was 49% (n = 173) in metastatic melanoma, 29% (n = 260) in advanced bladder cancer and 64% (n = 469) in EGFR+ aNSCLC (Table 2 & Supplementary Table 3). The proportion of patients who were subsequently treated with a future AA drug following treatment with an initial drug was 20% (70/353) in metastatic melanoma, 7% (65/888) in advanced bladder cancer and 22% (158/734) in EGFR+ aNSCLC. For bladder cancer, the median OS for the initial treatment was lower (37.6 months compared with 57.1 and 56.2 months for EGFR+ aNSCLC and metastatic melanoma, respectively), leading to a reduced likelihood that patients would remain alive to access new drugs for the next line of therapy.

Mean survival gains of AA drugs in the real-world database

The difference in mean survival for patients treated with the initial drug and then treated with the AA drug compared with SoC/no treatment was highest for the CITs in second-line and beyond metastatic melanoma (31.3 months), followed by osimertinib in EGFR+ aNSCLC (24.3 months), and third-line and beyond enfortumab vedotin-ejfv in advanced bladder cancer (16.3 months; Table 2 & Supplementary Figure 1). In patients treated with the initial drug, the average additional survival gain as a result of accelerated access to future drugs via the AA pathway was 7.0 months in metastatic melanoma, 5.2 months in EGFR+ aNSCLC and 1.2 months in advanced bladder cancer.

Receipt of AA drugs in the US population

After rescaling patient-level estimates from the real-world database to reflect the utilization of drugs within each tumor in the USA, the proportion of patients projected to survive to the AA date of the next innovation was 40% (n = 1743) in metastatic melanoma, 31% (n = 6635) in advanced bladder cancer and 62% (n = 12,646) in EGFR+ aNSCLC (Table 3 & Supplementary Table 3). The proportion and number of initial drug patients at the US population level treated with a subsequent future AA drug was 17% (745/4394) in metastatic melanoma, 8% (1698/21 488) in advanced bladder cancer and 21% (4233/20 317) in EGFR+ aNSCLC.

Mean survival gains attributable to the AA pathway in the US population

In the primary counterfactual analysis, where the traditional approval for the AA drug would have occurred at the time of conversion from AA approval to full approval, the number of initial drug patients subsequently receiving the AA drug owing to the AA pathway in the USA was 523/745 in metastatic melanoma, 1247/1698 in advanced bladder cancer and 2430/4233 in EGFR+ aNSCLC (Table 3). The AA pathway was estimated to contribute 3.73 additional months of survival per patient among all patients who initiated ipilimumab in metastatic melanoma (in the pre-AA period), 0.85 months per patient among all patients initiating CITs post platinum therapy in advanced bladder cancer and 2.89 months per patient among all patients initiating first- and second-generation EGFR TKIs in EGFR+ aNSCLC. These survival gains attributable to the AA pathway were 70, 73 and 57% of the total average survival gains across all initial drug patients, respectively, and represent a 3.4-, 3.8- and 2.3-times increase in survival compared with the traditional approval pathway. Extrapolating the attributable survival gains to the US population resulted in an estimated 1367 (metastatic melanoma), 1702 (advanced bladder cancer) and 4930 (EGFR+ aNSCLC) LYs gained owing to the AA pathway.

In scenario analyses, assuming a shorter time to traditional approval (i.e., date of clinical data presentation) resulted in survival gains attributable to the AA pathway ranging from 46 to 61% and LYs gained ranging from 827 to 38,934. In contrast, assuming a longer time to traditional approval (i.e., average oncology conversion time) increased the proportion of survival gains attributable to the AA pathway ranging from 82 to 92% and the LY gains ranging from 1772 to 7075.

Outcomes associated with the AA pathway varied by the timing of the initial drug start. The number of patients receiving a future AA drug (as a result of the AA pathway) and the average additional survival gains due to the AA pathway was largest for patients who initiated their initial drug closest to the AA approval date (Figure 2A & B). The number of LYs gained owing to the AA pathway was also largest for those initiating the initial drug closer to the AA approval date (Figure 2C).

Discussion

In this study using real-world data, we found that the AA pathway played a key role in enabling access to future AA drugs for patients who initiated a prior line of therapy in three case studies for solid tumors. Specifically, we found that the majority of the additional survival gains added to the initial drug from use of the future AA drug was attributable to the AA pathway, resulting in substantial LYs gained at a US population level. To our knowledge, this is the first study estimating the ROV of the AA pathway by measuring the real-world survival benefits of earlier access to oncology drugs among patients already treated with existing therapies.

Our analysis offers a more complete understanding of the benefits and consequences of the AA program via evidence of a potential benefit of the AA pathway previously not considered. Previous research using modeling has suggested that drugs approved via expedited approval pathways have greater health outcome gains than drugs approved via traditional approval pathways [22]. Our study is complementary to these previous findings because our analyses demonstrate that the AA pathway can shorten the time between innovations thereby increasing the potential of a larger proportion of patients to access sequential innovative therapies to provide additional real-world survival gains. An important and differentiating point is that this additional survival value accrues to the initial drug – the life extension due to the initial drug is improved by the AA pathway for a future innovative drug. Given this observation, the ability of the AA pathway to shorten the time between innovations should be considered when deliberating potential modifications to the AA program.

Considering that only a proportion of patients on the additional drug remained alive to access the future AA drug, the LYs gained estimates are substantial. For metastatic melanoma, 725 patients remained alive to access the future AA drug, leading to 1367 more population LYs gained over only 15.7 months (the conversion time for the primary analysis). Similarly, future AA drug access for 1698 patients led to 1702 LY gains in 19 months for advanced bladder cancer, and future AA drug access for 4233 patients led to 4930 LYs gained over 16.8 months in EFGR+ aNSCLC. As the AA pathway provides a one-time increase in early access, these population estimates of LYs gained translate into meaningful differences for individual patients facing limited survival on current treatment options. The combined effect of an initial treatment with extended survival, as well as the ability to increase access to future AA drug(s), led to OS estimates that were 1.6 (metastatic melanoma) to 2.7 (EGFR+ aNSCLC) times as long as survival for patients who did not have access to future AA drugs.

While overall we found that the AA pathway contributes meaningfully to survival gains, there was variability in the magnitude of the benefit depending on the proximity in time of patients initiated on the initial drug therapy to the AA date. In general, those patients who initiated their initial drug closer to the AA date of the future drug had the most benefit from the AA pathway. This observation is driven by both the effectiveness of the AA drug as well as the severity of the tumor and suggests that patients with tumors with rapid disease progression may stand to benefit more from shortening the time between innovations via the AA pathway. This observation also highlights the importance of rapid innovation and facilitating access to these therapies in areas of high unmet need (such as limited survival) because delays could result in substantial LYs lost.

This study has a number of limitations. First, the primary analysis was based on an assumption, namely that approval via a traditional pathway did not occur. We addressed this limitation by varying the date of traditional approval in sensitivity analyses to explore a range of real-world timings for conversion based on potential earlier use when positive data are presented at clinical conferences and potential later use based on average conversion time in oncology. Across these scenarios, we consistently found that the counterfactual scenario still resulted in LYs gained attributable to the AA pathway. Second, these results may not be generalizable to all tumors with AA drugs. The benefit of shortening the time between innovations and increasing the probability of patients benefiting from future therapies requires specific situations where innovations are both rapid and in subsequent lines of therapy, thus the value contribution of AA should be considered in this context. This potential benefit of the AA program is not present for all approvals and is much more likely to be relevant in areas of rapid innovation where a number of drugs are in clinical development for an indication. Further, the three AA case studies examined in this analysis were for drugs that converted to traditional approval, and our sample did not include AA drugs that were withdrawn. We therefore presented the impact of the AA pathway on patient survival and LYs for a subset of AA drugs. However, only about 10% of all products, and 8% of all oncology products, in the AA pathway have been withdrawn to date [26,27]. Additionally, other health outcomes, such as patients' quality of life and the cost impact of AA products on spending, were not measured in this analysis. Future research would be required to understand whether the benefit of bridging between innovations manifests in other health outcome measures. Finally, this analysis sought to isolate the impact of the AA program on ROV, or situations where patients on current life-extending therapies use newly approved medicines as a next line of treatment. This research topic was chosen given the demonstrated increasing importance of this information in drug value assessment and clinical decision-making [6–8]. Given this focus, our analysis and commentary did not comprehensively discuss the current debate on the benefits and challenges of the AA program [28,29].

Conclusion

The AA pathway in oncology can increase the likelihood for patients to benefit from future drugs by shortening the time between innovative therapies. While research has examined the impact of AA drugs on patients after AA approval date, our research suggests that survival gains may also be realized for patients who start on therapy in earlier lines of treatment before the AA approval date. Given the finding that a majority of the additional survival gains due to the AA drug was attributable to the AA program and its impact on earlier access, this aspect of value of the AA pathway should be addressed when considering future modifications to the FDA's AA program.