Strategic risk assessment in oncology: Utilizing single-agent activity to boost combination therapy approvals

Abstract

Background

This study examines whether single-agent activity in early clinical phases correlates with the approval likelihood of combination therapies in oncology. Using the Intelligencia AI database, we investigated the impact of monotherapy efficacy on the success rates of combination therapies.

Methods

The analysis included combinational therapies across various solid tumor types, assessing the approval rates and the presence of monotherapy efficacy with cut-off date September 12, 2024, and an assessment of 3896 programs. We analyzed historical clinical trial data focusing on Objective Response Rate (ORR) as a metric of single-agent activity (SAA).

Results

Approval rate for combination programs across all indications and phases was 4.2 %. Programs that included approved monotherapy drugs had an approval rate of 6.1 %, whereas those without approved monotherapy drugs had a lower approval rate of 2.7 %. However, the historical approval rate for combination programs with failed monotherapy drug with more than 20 % objective response rate was 5.8 %. Furthermore, approved combinations derived from monotherapy-failed pipelines showed diverse ORR thresholds, with specific trends observed across different cancer types.

Conclusions

The likelihood of approval for combination therapies is higher when combined with monotherapy drugs that have previously shown single agent activity. This finding is consistent with other research on historical approval rates and the common consensus within oncology drug development. Here we suggest that by leveraging monotherapy drug activity there can be an enhanced prioritization of anti-cancer agents repurposed for combination therapies, which would have otherwise been shelved based on their single agent failure.

Highlights

• •Drugs with single agent activity or demonstrable anti-tumor tumor activity as a monotherapy retains the highest chance of a successful combination therapy. Although this is a common view in drug development, this paper provides evidence in support of this paradigm from past trials.
• •The paper advocates strategic trial designs that integrate single-agent activity, thereby optimising resource allocation and accelerating drug development in oncology.
• •In exceptional cases, drugs that initially fail as monotherapies can contribute to successful combination therapies if they target distinct pathways or biomarkers. This finding is consistent with other published results and practice in drug development.

1. Introduction

The quest to enhance risk mitigation in oncology drug development is critical to provide patients with improved therapeutic options while managing the investment budgets of drug developers. With an estimated $1.1 billion investment per asset [1], late-stage failures are particularly costly, both financially and in terms of patient outcomes. These failures not only result in direct financial losses but also represent significant opportunity costs, including the time lost—often a decade or more—in pursuing unsuccessful therapies. This delay hinders the allocation of resources to potentially more promising projects and deprives patients of timely access to new treatments. Approval rates in oncology between 2000 and 2025 were estimated around 3.4 % significantly lower than in other therapeutic areas, underscoring the importance of improving decision-making processes throughout the drug development lifecycle [[2], [3], [4]].

Combination therapies have emerged as a cornerstone of oncology treatment. Combination approval rates in some analyses are reported to be as high as approximately 12 % if they complete a successful phase 1 study. While 40 % of combinations with evidence of activity at acceptable levels of toxicity in phase 1 progress further, only 5 % without such results do [5].

Recent examples, such as the approval of Relatlimab with Nivolumab and Capivasertib with Fulvestrant, illustrate cases where combination approvals were achieved despite limited monotherapy activity [[6], [7], [8]]. This study delves into the historical approval data, seeking to assess correlation between monotherapy efficacy and combination therapy approval, thereby informing future trial designs and reducing attrition rates in oncology.

2. Methods

2.1. Data sources

The study leveraged Intelligencia AI proprietary database, which undergoes routine updates and contains detailed, comprehensive information on industry-driven non-observational studies with FDA regulatory intent. For this research, we systematically retrieved historical data in the field of Oncology from January 1, 2000, onwards. The data analysis cut-off date for this investigation was September 12, 2024.

2.2. Program identification

Historical, industry-sponsored programs seeking FDA approval, and involving active interventions (e.g., non-observational clinical trials) utilizing a singular pharmaceutical agent were identified within the Intelligencia AI database from January 1, 2000, onwards. Within Intelligencia AI, a program is defined as one or more clinical trials illustrating the evolution of a specific drug or drug combination within a distinct disease context and specific patient population. Variations in drug dosage, administration mode, adjuvant setting, biomarker information, line of treatment (LoT), disease stage, primary sponsor, indication (e.g., Breast Carcinoma), or sub-indication (e.g., Her 2+ Breast Cancer) delineate program uniqueness (Fig. 1). Each program traces a drug or drug combination through clinical phases until regulatory outcomes are achieved, be it success, failure, or termination. A typical program may include multiple trials, and a single trial can contribute to more than one program. Monotherapy regimens employed a single pharmaceutical agent without concurrent administration of other drugs or interventional modalities like radiotherapy, surgery or androgen deprivation therapy. Combinational regimens included drug combinations or drug-intervention combinations. Programs incorporating specific drug formulations were categorized as monotherapy programs, similar to those integrating co-formulations, resulting in the development of distinct and novel pharmaceutical entities.

Fig. 1.

Program definition and diagram delineating ORR selection for the monotherapy pipelines.

2.3. Indication selection and data analysis

The analysis focused on pivotal solid cancer indications, including some of the most prevalent malignancies such as Bladder Cancer, Kidney Cancer, Esophageal Cancer, Brain Cancer, Lung Cancer, Prostate Cancer, Melanoma, Liver Cancer, and Breast Cancer, along with their sub-indications. Consort diagram in Fig. 2 outlines the steps taken to obtain the programs used for the analysis. Liquid cancers were excluded from the scope of the analysis, mainly due to the heterogeneity of how responses are reported. Historical programs are defined as the ones that have either received regulatory approval by the FDA, or the ones that have been discontinued for any reason. Ongoing programs are defined as the ones that are still in active clinical development and there is still a possibility of future regulatory approval by the FDA. Only trials that have a clear specification of Phase are included in the Intelligencia database. Notably, FDA-accelerated approvals are considered equivalent to regular approvals in this analysis. Monotherapy programs lacking regulatory approvals were classified based on their objective response rate (ORR) into activity brackets as shown in Table 1. In this setting, regulatory approval refers exclusively to oncology indications. Focusing on prevalent solid cancer indications like breast, lung, prostate, and liver cancers is critical because they represent the bulk of global cancer cases, accounting for approximately 3/4 of solid tumor incidences. These cancers, including their subtypes, are major contributors to cancer-related morbidity and mortality worldwide, reflecting both their high prevalence and the substantial unmet medical need for more effective treatments. For example, lung and breast cancers alone make up over 20 % of new global cancer diagnoses [9,10].

Fig. 2.

Consort diagram to estimate historical approval rate for combination programs with and without regulatory approval as a single agent and to determine single agent activity in monotherapy programs.

Table 1.

Definition of single agent activity at program level using objective response rate (ORR).

Single Agent Activity Level	Category of ORR
Robust activity	ORR >20 %
Moderate activity	10 % ≤ ORR ≤20 %
Limited activity	0 % < ORR <10 %
No activity	0 %
Unknown	ORR was not measure or not an endpoint

Solid cancers dominate the oncology landscape due to their prevalence and the complexity of managing these conditions with a combination of surgery, chemotherapy, immunotherapy, and other modalities. This makes them a focus for pharmaceutical companies and clinical trials. Data were analyzed using descriptive statistics based on the proportion of approved programs out of a total number of programs defined by the different sets of variables, including the single agent activity categories. A Chi-square test was used to test for significant association at 5 % level of significance. No adjustment for multiple testing is considered to keep the results exploratory.

2.4. Determination of single agent activity and ORR selection

We started our analysis by gathering single agent activity data from each trial at the cohort level. A cohort is defined as a subgroup of patients associated with an arm (e.g., key drug), characterized by variations in factors such as age, prior treatments, biomarker expression, smoking status, and other relevant parameters. The cohort-level ORR is extended to the trial level by comparing various ORR values and across multiple trials within a program. The optimal ORR for the cohort is selected as the highest value, considering the similarity of time frames across different trials for each indication, phase, and overall. Similarly, ORR at the trial level is extrapolated to the program level by identifying the maximum ORR at the trial level, accounting for variations in trial timeframes. For instance, if a program presents ORR from both dose escalation and dose expansion studies, the ORR from the dose expansion arm is prioritized due to its pivotal role in further clinical development of an asset as seen in Fig. 1.

To determine the ORR per program, non-pivotal trial cohorts, cohorts without declared patients, and escalation cohorts (if the corresponding expansion cohorts were unavailable) were excluded. If ORR data was unavailable, the sum of Complete Response Rate (CRR) and Partial Response Rate (PRR) was used instead (Fig. 1).

We categorized drugs as ‘approved’ if they had received FDA approval in the tested indication. Assets were included in the analysis only if they had never achieved approval as a monotherapy in the indications evaluated in this study. Conversely, if a program did not secure approval as single agents, the determination of SAA relied on the ORR, as outlined previously. The categorized were defined as no activity, limited activity, moderate and robust activity with ORR as shown in Table 1. The best objective response rate from all cohorts within the program was extracted to determine program-level SAA.

3. Results

A total of 2213 combinational and 1683 monotherapy historical (failed or approved) programs were extracted from the Intelligencia database, providing a comprehensive dataset for analyzing approval rates and identifying trends in oncology drug development.

3.1. Historical approval rates for combinational therapy

Between 2017 and 2022, the US Food and Drug Administration (FDA) approved 161 new cancer approvals for various solid tumor malignancies. In contrast, during the previous 13 years (2002–2014), only 71 treatments were approved for metastatic, advanced, and/or refractory solid cancers [11,12].

This surge in cancer therapies has led to increased treatment options and regulatory complexity. The cumulative approval rate of combinational therapies across all indications and clinical phase I-III was 4.2 %, equating to 93 approved programs out of 2213 (Table 2). Interestingly for phase III programs the overall percentage rises to 24 % overall and to 30.8 % and 15.7 % for approved and failed monotherapy-deriving programs respectively (Table S3 in the supplementary materials). Table S5 in the supplementary material shows approval rates for specific drug combinations with a combination involving immunotherapy and chemotherapy having 0.9 % and 5.5 % approval rates, respectively. To understand the factors influencing these approval rates, we examined the data by association with the success or failure of the combination program's corresponding monotherapy results, clinical phase and by specific cancer indications. Programs incorporating previously approved monotherapy (in oncology) drugs into combination therapies had an overall approval rate of 6.1 %, based on 55 approved combinations out of 907 total programs. This is higher than the general approval rate for all combinational therapies. In contrast, combination programs involving monotherapy drugs that had not secured FDA approval had a lower overall approval rate of 2.7 % based on 23 approved combinations out of 856 total programs. This comparison between approved and non-approved monotherapy drugs in combination therapies reveals the significant impact of prior approval status on subsequent success rates in combination regimens. Note that Approved monotherapy refers exclusively to oncology indications and no other therapeutic indications were considered.

Table 2.

Cumulative approval rate across indications and across all combinational programs irrespective of the phase of each included trial (${\chi }_1^2=11.09,p-value=0.0009$).

	Phase	Total Combos	Approved Combos	Approval Rate
overall	Overall	2213	93	4.20 %
approved mono	Overall	907	55	6.10 %
failed mono	Overall	856	23	2.70 %

Table 3 presents approval rates for specific cancer indications with varying success rates. For example, breast cancer had an approval rate of 5.5 % across phase I – III, based on 28 approved combinations out of 506 total programs, while esophageal cancer stood out with a 11.1 % approval rate based on 3 approved combinations out of 27 total programs. In contrast, brain cancer had a notably lower approval rate of 1.3 %, based on only 2 approved combinations out of 151 total programs. This indication-specific analysis underscores the variability in drug development success depending on the cancer type, suggesting that some indications may present greater challenges or opportunities for combinational therapies. By combining phase-specific and indication-specific insights, we can better comprehend the dynamics that influence the overall success of combination oncology therapies.

Table 3.

Cumulative approval rate per indication and across all combinational programs irrespective of the phase of each included trial (${\chi }_9^2=18.70,p-value=0.0279$).

Indication	Total Combos	Approved Combos	Approval Rate
Bladder cancer	104	4	3.80 %
Brain cancer	151	2	1.30 %
Breast cancer	506	28	5.50 %
Esophageal cancer	27	3	11.10 %
Kidney cancer	115	6	5.20 %
Liver cancer	111	3	2.70 %
Lung cancer	622	20	3.20 %
Prostate cancer	208	16	7.70 %
Skin cancer & melanoma	216	6	2.80 %
Stomach cancer	153	5	3.30 %

3.2. Comparison of successful combination programs stemming from failed monotherapy drugs and evaluation of SAA by phase and indication

SAA, categorized by ORR levels, was analyzed to predict combination therapy success, with ORR >20 % achieving the highest approval rate of 5.8 % (12 approvals out of 207 programs), followed by 5.1 % (8 approvals out of 156 programs with ORR 10–20 %) and 2.0 % (5 approvals out of 254 programs with ORR <10 %) (Table 4). Phase-specific analysis showed that combinations with robust SAA (>20 %) had approval rates of 3.4 % in phase 1, 6 % in phase 2, and 20.5 % in phase 3. Similarly, moderate activity in failed monotherapies led to approval rates of 6.4 % in phase 1, 4.8 % in phase 2, and 33.3 % in phase 3. Among 23 approved combinations involving failed monotherapies (Table 5), 12 had robust activity, 6 had moderate activity, and 5 displayed limited activity, with no approvals for combinations with monotherapy ORR of 0 %, consistent with findings by Yamaguchi and Walker et al. [2,13].

Table 4.

Overall combination approval rates based on not-approved monotherapy single agent activity (${\chi }_5^2=18.78,p-value=0.0021$).

Single Agent Activity	Category of ORR	Total Combinations	Approved Combinations	Approval Rate
Robust Activity	ORR >20 %	207	12	5.80 %
Moderate Activity	10 % ≤ ORR ≤20 %	156	8	5.10 %
Limited Activity	0 % < ORR <10 %	254	5	2.00 %
No Single Agent Activity	0 %	70	0	0.00 %
Unknown Activity		54	0	0.00 %
Not Measured		157	0	0.00 %

Table 5.

Summary of successful combination programs stemming from failed monotherapy drugs.

Indication	Primary drug	Additional drug	Primary MOA	Additional MOA	SAA bracket	%ORR value	Primary drug sponsor
Carcinoma, renal cell	Bavencio	Inlyta	Antagonist Programmed cell death 1 ligand 1 Protein	Inhibitor Vascular endothelial growth factor receptor 1 Protein, Inhibitor Vascular endothelial growth factor receptor 2 Protein, Inhibitor Vascular endothelial growth factor receptor 3 Protein	Activity	16.1
Gastric carcinoma	Herceptin	Cisplatin, Unspecified chemotherapy	Antagonist Receptor tyrosine-protein kinase erbB-2 Protein	Cross-Linking Agent DNA Nucleic-acid, Other Undefined Target Undefined Target	Robust Activity	33.3	Genentech, Hoffmann-La Roche
HER2 positive breast carcinoma	Perjeta	Docetaxel, Herceptin	Antagonist Receptor tyrosine-protein kinase erbB-2 Protein	Inhibitor Tubulin Protein complex group, Antagonist Receptor tyrosine-protein kinase erbB-2 Protein	Limited Activity	9.52	Genentech, Hoffmann-La Roche
HER2 positive breast carcinoma	Perjeta	Herceptin, Unspecified chemotherapy	Antagonist Receptor tyrosine-protein kinase erbB-2 Protein	Antagonist Receptor tyrosine-protein kinase erbB-2 Protein, Other Undefined Target Undefined Target	Limited Activity	9.52	Genentech, Hoffmann-La Roche
Non-Small cell lung carcinoma	Tafinlar	Mekinist	Inhibitor RAF proto-oncogene serine/threonine-protein kinase Protein, Inhibitor Serine/threonine-protein kinase B-raf Protein, Inhibitor Serine/threonine-protein kinase B-raf V600D Mutated protein, Inhibitor Serine/threonine-protein kinase B-raf V600E Mutated protein, Inhibitor Serine/threonine-protein kinase B-raf V600K Mutated protein	Reversible inhibitor Dual specificity mitogen-activated protein kinase kinase 1 Protein, Reversible inhibitor Dual specificity mitogen-activated protein kinase kinase 2 Protein	Robust Activity	34.6	Novartis
Non-Small cell lung carcinoma	Abraxane	Paraplatin	Inhibitor Tubulin Protein complex group	Cross-Linking Agent DNA Nucleic acid	Limited Activity	9.6	ABRAXIS BIOSCIENCE, Celgene
HER2 positive breast carcinoma	Lapatinib	Xeloda	Inhibitor Epidermal growth factor receptor Protein, Inhibitor Receptor tyrosine-protein kinase erbB-2 Protein	Disrupting Agent DNA Nucleic-acid, Disrupting Agent RNA Nucleic-acid, Inhibitor Thymidylate synthase Protein	Robust Activity	39	NATCO Pharma, Novartis
Melanoma	Braftovi	Mektovi	Inhibitor RAF proto-oncogene serine/threonine-protein kinase Protein, Inhibitor Serine/threonine-protein kinase B-raf Protein, Inhibitor Serine/threonine-protein kinase B-raf V600E Mutated protein	Inhibitor Dual specificity mitogen-activated protein kinase kinase 1 Protein, Inhibitor Dual specificity mitogen-activated protein kinase kinase 2 Protein	Robust Activity	60	Array BioPharma
HER2 positive breast carcinoma	Lapatinib	Femara	Inhibitor Epidermal growth factor receptor Protein, Inhibitor Receptor tyrosine-protein kinase erbB-2 Protein	Inhibitor Aromatase Protein	Robust Activity	39	NATCO Pharma, Novartis
Non-Squamous non-small cell lung cancer	Tecentriq	Abraxane, Paraplatin	Antagonist Programmed cell death 1 ligand 1 Protein	Cross-Linking Agent DNA Nucleic-acid, Inhibitor Tubulin Protein complex group	Limited Activity	9.6	Genentech
Melanoma	Tecentriq	Cotellic, Zelboraf	Antagonist Programmed cell death 1 ligand 1 Protein	Inhibitor Dual specificity mitogen-activated protein kinase kinase 1 Protein, Inhibitor Dual specificity mitogen-activated protein kinase kinase 2 Protein, Inhibitor Serine/threonine-protein kinase B-raf Protein, Inhibitor Serine/threonine-protein kinase B-raf V600A Mutated protein, Inhibitor Serine/threonine-protein kinase B-raf V600D Mutated protein, Inhibitor Serine/threonine-protein kinase B-raf V600E Mutated protein, Inhibitor Serine/threonine-protein kinase B-raf V600G Mutated protein, Inhibitor Serine/threonine-protein kinase B-raf V600K Mutated protein, Inhibitor Serine/threonine-protein kinase B-raf V600M Mutated protein	Robust Activity	38.5	Genentech
Clear cell renal carcinoma	Lenvatinib	Everolimus	Inhibitor Fibroblast growth factor receptor 1 Protein, Inhibitor Fibroblast growth factor receptor 2 Protein, Inhibitor Fibroblast growth factor receptor 3 Protein, Inhibitor Fibroblast growth factor receptor 4 Protein, Inhibitor Mast/stem cell growth factor receptor Kit Protein, Inhibitor Platelet-derived growth factor receptor alpha Protein, Inhibitor Proto-oncogene tyrosine-protein kinase receptor Ret Protein, Inhibitor Vascular endothelial growth factor receptor 1 Protein, Inhibitor Vascular endothelial growth factor receptor 2 Protein, Inhibitor Vascular endothelial growth factor receptor 3 Protein	Inhibitor Serine/threonine-protein kinase mTOR Protein, Modulator Peptidyl-prolyl cis-trans isomerase FKBP1A Protein	Robust Activity	27	Eisai
Carcinoma, renal cell	Lenvatinib	Keytruda	Inhibitor Fibroblast growth factor receptor 1 Protein, Inhibitor Fibroblast growth factor receptor 2 Protein, Inhibitor Fibroblast growth factor receptor 3 Protein, Inhibitor Fibroblast growth factor receptor 4 Protein, Inhibitor Mast/stem cell growth factor receptor Kit Protein, Inhibitor Platelet-derived growth factor receptor alpha Protein, Inhibitor Proto-oncogene tyrosine-protein kinase receptor Ret Protein, Inhibitor Vascular endothelial growth factor receptor 1 Protein, Inhibitor Vascular endothelial growth factor receptor 2 Protein, Inhibitor Vascular endothelial growth factor receptor 3 Protein	Antagonist Programmed cell death protein 1 Protein	Robust Activity	27	Eisai
Non-Muscle-Invasive bladder cancer	Nogapendekin alfa	BCG vaccine	Agonist Cytokine receptor common subunit gamma Protein, Agonist Interleukin-2 receptor subunit beta Protein	Immunostimulant Undefined Target Undefined Target	Activity	20
Triple negative breast neoplasms	Keytruda	Unspecified chemotherapy	Antagonist Programmed cell death protein 1 Protein	Other Undefined Target Undefined Target	Activity	15.6	Merck Sharp & Dohme
HER2 positive breast carcinoma	Margenza	Unspecified chemotherapy	Antagonist Receptor tyrosine-protein kinase erbB-2 Protein	Other Undefined Target Undefined Target	Activity	12	MacroGenics
Non-Small cell lung carcinoma	Braftovi	Mektovi	Inhibitor RAF proto-oncogene serine/threonine-protein kinase Protein, Inhibitor Serine/threonine-protein kinase B-raf Protein, Inhibitor Serine/threonine-protein kinase B-raf V600E Mutated protein	Inhibitor Dual specificity mitogen-activated protein kinase kinase 1 Protein, Inhibitor Dual specificity mitogen-activated protein kinase kinase 2 Protein	Limited Activity	9.1	Array BioPharma
Gastric carcinoma	Keytruda	Herceptin, Unspecified chemotherapy	Antagonist Programmed cell death protein 1 Protein	Antagonist Receptor tyrosine-protein kinase erbB-2 Protein, Other Undefined Target Undefined Target	Robust Activity	33.3	Merck Sharp & Dohme
Gastric carcinoma	Keytruda	Unspecified chemotherapy	Antagonist Programmed cell death protein 1 Protein	Other Undefined Target Undefined Target	Robust Activity	33	Merck Sharp & Dohme
Carcinoma, hepatocellular	Tecentriq	Avastin	Antagonist Programmed cell death 1 ligand 1 Protein	Inhibitor Vascular endothelial growth factor A, long form Protein	Activity	17	Genentech
Carcinoma, hepatocellular	Opdivo	Yervoy	Antagonist Programmed cell death protein 1 Protein	Antagonist Cytotoxic T-lymphocyte protein 4 Protein	Robust Activity	23	Bristol-Myers Squibb
Low grade glioma	Tafinlar	Mekinist	Inhibitor RAF proto-oncogene serine/threonine-protein kinase Protein, Inhibitor Serine/threonine-protein kinase B-raf Protein, Inhibitor Serine/threonine-protein kinase B-raf V600D Mutated protein, Inhibitor Serine/threonine-protein kinase B-raf V600E Mutated protein, Inhibitor Serine/threonine-protein kinase B-raf V600K Mutated protein	Reversible inhibitor Dual specificity mitogen-activated protein kinase kinase 1 Protein, Reversible inhibitor Dual specificity mitogen-activated protein kinase kinase 2 Protein	Robust Activity	70.6	Novartis
Triple negative breast neoplasms	Keytruda	Unspecified chemotherapy	Antagonist Programmed cell death protein 1 Protein	Other Undefined Target Undefined Target	Activity	15.6	Merck Sharp & Dohme

Acronyms: MOA-mechanism of action. SAA – single agent activities; ORR – objective response rate.

Successful combinations predominantly targeted specific cancer types like breast, kidney, and hepatocellular carcinoma, utilizing complementary mechanisms such as erb-2 antagonists with chemotherapy or VEGFR and FGFR inhibitors for angiogenesis disruption. Combinations developed within a single company (17.4 % of cases) benefited from internal data, enabling optimized trial designs and efficient progress through testing and approval, highlighting a strategic advantage.

The analysis demonstrated that robust SAA correlates with increased combination therapy success, particularly in later-phase trials, emphasizing the importance of defining SAA thresholds for failed monotherapy agents. In skin cancer (Table S1 in the supplementary materials), robust SAA led to a combination approval rate of 6.9 %, outperforming monotherapy approval rates of 4.7 % (Table S2 in the supplementary materials), while limited activity (ORR <10 %) resulted in no approvals. Among the Breast cancer programs, 26.98 % had no no SAA, 34.13 % had limited activity, 15.87 % had moderate activity, and 23.02 % had robust activity (Table S4 in the supplementary materials). Successful combinations included HER2-positive breast cancer (e.g., Margetuximab with chemotherapy, Lapatinib with Capecitabine and Letrozole) and melanoma (Encorafenib with Binimetinib), where SAA thresholds of 10–20 % or higher were exceeded, leveraging biomarker-targeted approaches.

4. Conclusion

It is critical to promptly identify promising combination therapies, along with the associated factors, if we are to improve approval rates in oncology. The paradigm shift towards combination therapy, based on the biological hypothesis that combining drugs targeting different or complementary biological pathways can provide added benefits [14,15], necessitates the question of whether presence or absence of monotherapy activity correlates with the likelihood of approval of a combination therapy.

This study highlights a significant connection between SAA observed in early clinical phases and the likelihood of approval for combination therapies in oncology. Using the Intelligencia AI database, we explored how monotherapy efficacy impacts approval rates across different cancer types. Results showed that the overall approval rate for combination programs across all indications and phases was 4.2 %, based on 93 approved combinations out of 2213 total combination programs. Programs that incorporated previously approved monotherapy drugs into combination therapies achieved an approval rate of 6.1 %. Conversely, programs involving monotherapy drugs that lacked FDA approval had a lower approval rate of 2.7 %. Our dataset also highlights the variability in approval rates across different cancer types, with indications such as breast, kidney, and skin cancers showing higher success rates. However, cancers like brain, prostate, and bladder demonstrate lower approval rates for combination therapies, reflecting distinct clinical and regulatory challenges in these fields. Our findings align with common knowledge within the drug development community and are consistent with Walker and Newell's emphasis on the importance of distinguishing between drug types, such as targeted therapies, and the role of biomarkers in drug development [2,15]. Incorporating agents without documented single-agent activity into a treatment regimen is unlikely to produce meaningful improvements in activity unless there is a compelling biological rationale [16,17].

Despite the insightful findings, this study has several limitations. Our analysis relies solely on the use of ORR/CRR/PRR as a readout of efficacy, without incorporating other critical metrics such as OS or PFS. OS and PFS are key indicators of a treatment's sustained effectiveness and impact on patient prognosis. By excluding these metrics, our analysis may not fully reflect the long-term benefits or limitations of the therapies evaluated, potentially leading to an incomplete understanding of their clinical value. Incorporating OS and PFS in future analyses would provide a more comprehensive assessment of therapeutic efficacy and its implications for patient care. Additionally, our study only covers selected solid cancers, excluding liquid cancers, which may limit the generalizability of the results across all oncology indications.

CRediT authorship contribution statement

Adetayo Kasim: Writing – review & editing, Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Conceptualization. Marina Anastasiou: Writing – review & editing, Writing – original draft, Visualization, Methodology, Formal analysis, Data curation. Evangelia Loizou: Writing – review & editing, Writing – original draft, Visualization, Validation, Project administration, Formal analysis, Data curation, Conceptualization. Andreas Dimakakos: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Methodology, Investigation, Formal analysis, Data curation. Maria Georganaki: Writing – review & editing, Writing – original draft, Supervision, Resources, Project administration. Ioannis Mourelatos: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Data curation. Panos Karelis: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Formal analysis, Data curation, Conceptualization. Marianna Esposito: Writing – review & editing, Writing – original draft, Visualization, Resources, Project administration. Helen Zhou: Writing – review & editing, Writing – original draft, Visualization, Resources, Funding acquisition, Conceptualization. Tai-Tsang Chen: Writing – review & editing, Writing – original draft, Visualization, Resources, Funding acquisition, Conceptualization. Dimitrios Skaltsas: Writing – review & editing, Writing – original draft, Visualization, Resources, Funding acquisition, Conceptualization. Paul Stockman: Writing – review & editing, Writing – original draft, Visualization, Supervision, Resources, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: At the time of the submission of this manuscript: Adetayo Kasim, and Paul Stockman are employee of GlaxoSmithKline, Stevenage, United KingdomHelen Zhou, Tai-Tsang Chen are employee of GlaxoSmithKline, Philadelphia, PA, United StatesMarina Anastasiou, Evangelia Loizou, Andreas Dimakakos, Maria Georganaki, Ioannis Mourelatos, Panos Karelis, Marianna Esposito and Dimitris Skaltas are employee of Intelligencia AI Inc, New York, NY 10014, USA.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

Data will be made available on request.