Regulatory Considerations for Contribution of Effect of Drugs Used in Combination Regimens: Renal Cell Cancer Case Studies

For applications to market a new drug or biologic, regulations require adequate and well-controlled studies that distinguish the effect of a drug from other influences (1). Drug development strategies vary based on the drug product under investigation, with specific considerations required for the development of combination regimens. Combination regimens may consist entirely of previously approved drugs, a novel agent in combination with a previously approved drug(s) or novel drugs in combination. The FDA has provided guidance for industry for the development of two or more new investigational drugs for use in combination to assist with drug development (2). This article, however, focuses on combinations of previously approved drugs.

The standard approach when developing combination drug regimens consists of a randomized controlled trial with a factorial design that includes a treatment arm for the combination, treatment arms for each monotherapy component in the combination regimen, and a control arm (if one of the drugs is not a standard-of-care option). This trial design provides an opportunity to evaluate the safety and efficacy of each agent individually, and in combination, compared with standard of care while also providing a clear understanding of the necessity of each drug in the combination (Fig. 1). An example of this trial design was used for the initial approval of oxaliplatin. The three arms in this trial consisted of the single drugs oxaliplatin, intravenous 5-FU, and a combination of these two drugs (a.k.a. FOLFOX4). The trial results demonstrated that the efficacy of single-agent oxaliplatin was similar to that of intravenous 5-FU/LV, and that oxaliplatin should not be used alone in this patient population (3). With monotherapy arms, this factorial trial design also allows for a comparison of the different components of the combination with respect to time-to-event endpoints such as progression-free survival (PFS) and overall survival (OS) as well as a more complete interpretation of combination regimen safety. An add-on trial is another approach employed in combination drug development. This approach is useful to demonstrate superiority of a combination of drugs when a new drug is added onto a standard-of-care backbone, such as was seen with the addition of CDK4/6 inhibitors to standard endocrine therapy in patients with hormone receptor–positive metastatic breast cancer (4–6).

To promote drug development in patients with cancer, some flexibility in trial design may be needed. Alternative approaches to combination drug trials may allow for accelerated development of combination regimens that are clearly superior to available therapy and could decrease patient exposure to less effective therapies. Alternative approaches may also be appropriate for drugs that have little to no effect as monotherapy but can potentiate the activity of another drug. Ultimately, drug applications must clearly demonstrate that all components of the combination regimen are necessary to produce the magnitude of the treatment effect while also showing that the clinical benefit of the combination is greater than the potential toxicity.

To overcome some of the limitations from cross-trial comparisons, a strong biologic rationale and nonclinical and/or early clinical evidence supporting the necessity of each drug in the proposed combination is critical. Preclinical studies, while helpful in drug development, do not predict for clinical activity of the drug and thus would be unlikely to supplant clinical information. Information from early clinical trials, such as from add-on trials, can support lack of activity of an individual drug. Strength of clinical evidence required to support the assessment of the contribution of each drug depends on the context of disease and population, availability and effectiveness of other treatments, and clinical data available for the individual drugs.

To determine the contribution of individual drugs to the effect of a combination regimen, FDA has typically relied on randomized data from the same trial that demonstrates the outcome of the monotherapies. Using external trial data of monotherapy drug activity to interpret its activity in combination treatment may be appropriate in situations as described above. However, this approach requires several assumptions such as a presumed similarity of the patient populations, absence of unidentified confounding variables in the supportive trials, and similar methods of response assessment and data collection across trials. In addition, the limitations of cross-trial comparison include determination of appropriate endpoints for comparison. Apparent differences in treatment effect may be the result of differences in factors related to the natural history of disease, imaging techniques used, response monitoring, and supportive care. For cross-trial comparisons, objective response rate (ORR) may be more appropriate than PFS or OS because tumor shrinkage would not be expected without intervention and therefore would reflect drug activity. However, use of ORR is not without potential issues. Differences in definitions of response, assessment method of response (e.g., investigator assessment or blinded independent central review), or the inability to confirm that objective response correlates to a clinically meaningful outcome with long-term impact on disease course may still limit comparisons across trials. In addition, prognostic characteristics are not accounted for as they would be with randomization. These limitations can make it difficult to draw conclusions on the contribution of each of the drugs across trials.

To overcome the uncertainties surrounding interpretation of cross-trial comparisons, it may be preferable to conduct a smaller randomized trial where contribution of effect can be more clearly delineated. Such a trial could be conducted earlier in development, designed as a smaller randomized trial with ORR as the primary endpoint to descriptively examine the necessity of each drug in a combination. Simon's 2-stage design could be incorporated into each monotherapy arm to allow for early stopping due to inactivity. This design allows for direct comparison of the combination regimen with treatment arms containing components of the combination. This approach can successfully achieve the desired information while enrolling fewer patients and limiting resource expenditure. Limitations of this approach include potential for inconsistencies of a smaller trial to predict long-term patient benefit with the combination compared with monotherapy as the trial would not be powered for detecting differences in PFS or OS. However, this approach allows drug development to proceed at its usual pace with a parallel or preliminary smaller trial assessing activity of the drugs with ORR as an endpoint.

Historical controls may allow for the use of large existing databases or literature resources in lieu of enrolling patients to concurrent control treatment arms. However, because standard of care and supportive management change over time, the data's age and relevance to current clinical practice are important for determining the utility of the historical control. In addition, inability to gather patient level data inhibits the ability to ensure the historical control is a representation of the intended patient population.

The potential role of real-world data in support of new drug applications is receiving increased attention. The 21st Century Cures Act, signed into law in 2016, is designed to help accelerate medical product development and has proposed the use of real-world data as one of the ways to modernize clinical trial design (7). The ease in which real-world data are used in the development of combination drug regimen will depend on the data source and quality, the age of the data and relevance to current clinical practice, and the type and quantity of missing data. Patient-level data may be unavailable when using both historical control data and real-world data. This limits the ability to adequately characterize baseline prognostic variables, appropriately match the characteristics of the control arm with the treatment arm, and to identify confounding variables in the control populations. Trial procedures and data collection practices may vary widely between historical controls, real-word data, and contemporary clinical trials, thus reducing the reliability of comparisons among the three.

In circumstances where baseline characteristics and/or disease status of patients on external control trials may differ from the characteristics of the patients on the trial assessing the investigational regimen, propensity score matching has been used to account for differences in patient populations and improve the utility of the external control for cross-trial comparisons (8). Although propensity score matching is a tool that has been used on an ad hoc basis by the FDA to manage variation in external control patient populations (as was done in the pembrolizumab/axitinib and avelumab/axitinib approvals), it does not account for all potential limitations associated with the use of external data to isolate the contribution of effect. A randomized control trial is still a preferred approach to isolate the contribution of effect of each drug in the combination regimen.

While using data from different sources, it is important to assess the difference in trial population, trial conduct, and endpoint assessment during trial design. Identifying potential biases and prespecifying methods to adjust for such biases are essential as our understanding of how to incorporate external controls evolves.

Adaptive trial designs also have potential to be a useful resource in the development of combination drug regimens. Adaptive randomization is a technique that allows the randomization of patients preferentially to treatment arms that are performing at a prespecified level of efficacy (9). A multi-arm adaptive trial design allows the collection of data on the contribution of effect of each agent, with the ability to drop one or more monotherapy arms if they are not meeting prespecified measures of efficacy. Both approaches have the potential to accelerate development of a combination drug regimen through less reliance on large sample sizes and the ability to limit unnecessary resource expenditure to clinical trial arms with little monotherapy efficacy, while still providing evidence of contribution of effect from a single randomized trial.

Between April 2018 and May 2019, the FDA granted marketing approvals for three combination drug regimens in previously untreated, locally advanced or metastatic renal cell carcinoma (RCC): nivolumab in combination with ipilimumab, pembrolizumab in combination with axitinib, and avelumab in combination with axitinib (10–14). The trials supporting these approvals used designs that compared the efficacy of the combination regimen against a standard-of-care control arm without including separate arms demonstrating the monotherapy efficacy of each agent in the combination regimen (Fig. 2). To demonstrate the contribution of each drug to the efficacy of the combination regimen, external clinical trial data demonstrating the monotherapy activity of the drugs was used. In these cases, each drug in the combination regimen was previously approved for oncology indications and had a well-known safety profile and previously reviewed efficacy data in other treatment settings. In contrast, other combination drug regimens for the treatment of advanced RCC have employed different trial design approaches including the combination of lenvatinib and everolimus, which utilized a three-arm factorial design and bevacizumab in combination with IFN, which utilized an add-on design (15–18).

Prior clinical experience with anti–PD-1 antibodies in combination with sunitinib or pazopanib demonstrated a high incidence of high-grade treatment-related adverse events (AE) and AEs leading to discontinuation, and this combination was considered inappropriate for further clinical development (19). For two of the trials, an anti–PD-(L)1 antibody was given, avelumab in one and pembrolizumab in the other trial, in combination with axitinib and compared with sunitinib monotherapy. In the case of nivolumab, prior clinical data of efficacy and safety existed to support the evaluation in combination with ipilimumab (20).

CheckMate 214 was the pivotal trial supporting approval of nivolumab in combination with ipilimumab in patients with intermediate or poor risk, previously untreated advanced renal cell carcinoma (21). The three co-primary endpoints were (i) Independent Radiological Review Committee–assessed ORR in intermediate or poor-risk patients; (ii) PFS in intermediate or poor-risk patients; and (iii) OS in intermediate or poor-risk patients, according to International Metastatic RCC Database Consortium (IMDC) prognostic scoring (22, 23). External data included trial CA209009, an immunomodulatory activity study of nivolumab monotherapy in the first-line metastatic RCC setting (Study CA209009; ref. 24), data from the nivolumab arm of the randomized second-line advanced RCC trial compared with everolimus (CheckMate 025; ref. 25), and monotherapy ipilimumab data from patients treated in the first-line metastatic RCC setting on MDX010-11. In the primary analysis for CheckMate 214, the ORR for the nivolumab plus ipilimumab combination [42%; 95% confidence interval (CI), 37–47] suggested improvement compared with the ORR from external trials of either nivolumab in the first or second line (1L 13%; 95% CI, 3–32; 2L 22%; 95% CI, 18–26; ref. 10) or ipilimumab alone (25%; 95% CI, 5–57) despite the overlap in CIs between the combination regimen and ipilimumab (Fig. 3). In addition to an improvement in ORR with the nivolumab plus ipilimumab combination, there was a clear improvement in OS in the intermediate or poor risk subgroup over the active control sunitinib (HR, 0.63; 95% CI, 0.44–0.89). This served as an important regulatory consideration for granting regular approval and helped overcome issues related to the trial design not directly isolating the effect of nivolumab or ipilimumab. Clinical evidence of efficacy of ipilimumab and nivolumab monotherapies compared with the combination regimen had been demonstrated, albeit at a different dosage regimen, in patients with treatment-naïve melanoma, which provided further support of the efficacy of the combination. In addition, the toxicity profile of nivolumab plus ipilimumab in RCC was considered acceptable, and the rates of grade 3–4 AEs were more common in the sunitinib arm than in the nivolumab plus ipilimumab arm (76% vs. 65%).

Scientific rationale for the combination of an anti–PD(L)-1 antibody with axitinib was based on the hypothesis that an antiangiogenic agent may help overcome immunotherapy resistance by eliminating physical barriers, such as tumor-associated vasculature, allowing immune cells to better infiltrate the tumor (26). KEYNOTE-426 randomized patients with treatment-naïve advanced RCC to pembrolizumab 200 mg every 3 weeks in combination with axitinib 5 mg twice daily versus sunitinib monotherapy at the approved dose and schedule (27). The two dual primary endpoints were OS and PFS. Patients were stratified by IMDC risk category and geographic region. Monotherapy data of pembrolizumab was submitted and reviewed from KEYNOTE-427, a single-arm, open-label trial evaluating pembrolizumab in the treatment of patients with first line locally advanced/metastatic RCC; ORR data from 110 patients with clear cell RCC from cohort A were submitted. Monotherapy axitinib data were reviewed from the axitinib arm of the randomized open label study of axitinib compared with sorafenib in the first-line treatment of patients with advanced RCC (28). In KEYNOTE-426, the ORR for pembrolizumab in combination with axitinib (59%; 95% CI, 54–64) was superior to ORRs from external trials of pembrolizumab (36%; 95% CI, 27–46) and axitinib (32%; 95% CI, 26–39) monotherapy with nonoverlapping CIs (Fig. 4). Propensity-score matching methods were applied to match patients from the combination therapy arm in KEYNOTE-426 to those from the pembrolizumab and axitinib monotherapy arms of the external trials. On the basis of the post hoc propensity-score matching analyses, contribution of effect was evaluated comparing combination therapy with monotherapies in the matched patient population. The ORRs observed for the combination and each monotherapy arm among the matched patients was similar to that observed in the overall population and supported the contribution of effect. The HR for the primary endpoint, OS, clearly demonstrated the superiority of pembrolizumab plus axitinib over sunitinib (0.53; 95% CI, 0.38–0.74), with 18% OS events out of all randomized patients.

The scientific rationale for combining avelumab with axitinib was similar to that of pembrolizumab. JAVELIN Renal 101 randomized patients with treatment-naïve advanced RCC to avelumab 10 mg/kg every 2 weeks in combination with axitinib 5 mg twice daily or sunitinib monotherapy at the approved dose and schedule (29). The two coprimary endpoints were OS and PFS. External trials of avelumab monotherapy from the phase Ib JAVELIN Solid Tumor trial (30), and the axitinib arm from the randomized open label study of axitinib compared with sorafenib in the first-line treatment of patients with advanced RCC were also submitted (28). In JAVELIN Renal, the ORR for avelumab in combination with axitinib (51%; 95% CI, 47–56) also demonstrated superiority over the avelumab monotherapy trial data (16%; 95% CI, 8–28) and the axitinib monotherapy data (32%; 95% CI, 26–39) with nonoverlapping CIs (Fig. 5). Propensity-score matching methods were also applied in this case. Among patients who were matched on propensity score in the post hoc analysis, the ORRs observed for the combination arm and each monotherapy arm among the matched patients was similar to that observed in the overall population and supported the contribution of effect. JAVELIN Renal 101 met the coprimary endpoint of PFS, with HR 0.69 (95% CI, 0.56–0.84), which formed the basis for approval. At the time of the application review, OS data were immature with 81 deaths observed (22% of information fraction of the total 368 deaths required for the preplanned final OS analysis); therefore, evidence of direct antitumor effect is based on ORR and evidence of clinical benefit is based on improvement in PFS supported approval.

The measures of overall safety of both the pembrolizumab plus axitinib and avelumab plus axitinib combinations, including number of deaths (2%–3% of patients within 30 days of treatment), serious adverse events (40% and 35%, respectively), and permanent treatment discontinuations due to adverse events (31% and 22%, respectively), were comparable with those of the sunitinib arms in the randomized trials.

The development of combination drug regimens has dramatically expanded the treatment landscape of oncologic disease and allowed improved patient outcomes compared with monotherapies. Although innovative approaches to expedite drug development are desired from all stakeholders, it is critical that patients are not exposed to ineffective drugs in a combination regimen that may increase toxicity with little additional clinical benefit or potential harm. There are examples of trials in multiple myeloma, where a reduced overall survival was reported when a drug was added to the combination (31). It is therefore important that combination therapy drug development programs clearly and objectively demonstrate that each agent in the combination is necessary for the treatment outcome.

Increasingly, this may occur through conduct of smaller randomized trials where contribution of effect can be more clearly delineated, utilizing endpoints such as ORR that can be measured earlier and are a definitive measure of drug activity, regardless of any cross-trial comparison, to overcome the uncertainties surrounding interpretation of cross-trial comparisons. Although certain situations may call for alternative approaches to drug development of combination therapy, these approaches are not without limitations. In the combination therapy approvals described above, the use of external clinical trial data to demonstrate the efficacy of the monotherapy drugs required cross-trial comparisons that introduce some uncertainty in the comparisons due to potential for heterogeneity between the two patient populations and differences in study procedures and response assessments. However, the increased ORR as a measure of definitive activity, the improvement in magnitude of a time-to-event endpoint such as OS observed with the combination regimen compared with monotherapy, prior FDA review of efficacy of the therapy particularly in the same disease and a strong biological rationale contributed in overcoming the uncertainty surrounding the use of external trial data. Approvals of multiple combination regimens for the same disease provides an advantage allowing for options in treatment of patients based on their risk factors and toxicity profile of the combinations.

Careful consideration regarding approach will allow for a better understanding of combination regimen safety and efficacy. The FDA encourages the discussion of development plans for combination drug regimens to ensure proposed trial designs will produce data that are adequate to attain a robust assessment and understanding of the combination regimen.

No potential conflicts of interest were disclosed.