Skip to main content

Estimands and the Patient Journey: Addressing the Right Question in Oncology Clinical Trials

Abstract

The diversity of patient journeys can raise fundamental questions regarding the evaluation of drug effects in clinical trials to inform clinical practice. When defining the treatment effect of interest in a trial, the researcher needs to account for events occurring after treatment initiation, such as the start of a new therapy, before observing the end point. We review the newly introduced estimand framework to structure discussions on the relationship between patient journeys and the treatment effect of interest in oncology trials. In 2017, the International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use released a draft addendum to its E9 guideline. The addendum introduces the concept of an estimand to precisely describe the treatment effect of interest. This estimand framework provides a structured approach to discuss how to account for intercurrent events that occur after random assignment and may affect the assessment or interpretation of the treatment effect. The framework is expected to improve coherence between trial objectives, design, analysis, and interpretation, as illustrated by examples in oncology disease settings. The estimand framework was applied to design a trial for a chimeric antigen receptor T-cell therapy. The treatment effect of interest was carefully defined considering the range of patient journeys expected for this particular indication and treatment. The trial design was developed accordingly to assess that treatment effect. All parties involved in the design of clinical trials need to consider possible patient journeys to define appropriate treatment effects and corresponding trial designs and analysis strategies. The estimand framework provides a common language to address the complexity introduced by varied patient journeys.

Introduction

A 64-year-old man presents to his oncologist with a persistent cough and a new diagnosis of adenocarcinoma of the lung. On the basis of the patient’s clinical and pathologic findings, he is offered entry into a randomized trial of a new drug plus standard-of-care therapy (SOC) versus SOC alone. Three months later, the patient has intolerable toxicity attributed to the treatment regimen and discontinues study treatment. The patient is started on a new anticancer regimen. A second patient presents with bloody sputum and is offered entry into the same randomized trial. After three months of treatment, the patient dies in a car accident.
In practice, a diversity of patient journeys is expected in oncology clinical trials. However, it is not clear how certain events that may occur within those patient journeys could affect the assessment of clinical benefit from the investigated treatment. The examples provided earlier include discontinuation of treatment as a result of an adverse reaction, start of a new anticancer regimen before observing the end point of interest (eg, progression-free survival [PFS] or overall survival [OS]), or death of a patient while on treatment.
Some of these events may reflect a desired or undesired activity of the treatment, which is not directly captured by the end point; some may prevent observation of the end point; and still others may complicate interpretation of the end point. Fundamental questions arise for the researcher. For example, when our focus is on PFS, should tolerability issues be reflected in the definition of PFS to measure clinical benefit? Is it still relevant and feasible to assess tumor progression after the patient has changed anticancer therapy? Which treatment effect would actually be assessed with the resulting PFS? Should death from a car accident be captured in the definition of PFS? Could that death indirectly reflect the patient’s health status or a treatment effect?
The range of journeys that a patient may experience, as illustrated in Figure 1, brings into the trial a number of events that occur after random assignment and possibly affect the assessment and interpretation of the treatment effect. Such events are termed intercurrent events.
Fig 1. Journeys of four patients illustrating the difficulty to define the treatment effect to be assessed between the two randomly assigned groups. The end point is progression-free survival (PFS), which is determined by assessing time to tumor progression or death. Solid dots at the end of the dashed line of the patient journey indicate that the assessments were performed and tumor progression was finally observed. Patient 1 discontinues treatment as a result of an adverse event, starts a different therapy, and remains in the efficacy follow-up. Patient 2 discontinues treatment and withdraws from further follow-up. Patient 3 dies in a car accident without prior progression, while still being on treatment. For those three patients, events occurring after random assignment either prevent the observation of the end point (patient 2: How long would the PFS be otherwise?) or raise important questions on its interpretation (patient 1: Is the observed PFS a result of the effect of the investigational treatment or of the new therapy? patient 2: Is the withdrawal of consent indicative of any treatment effect? patient 3: Is death a random event, or is it indirectly related to the disease or treatment?). By contrast, patient 4 has no intercurrent events affecting the observation or interpretation of the measured end point. EOT, end of treatment; SOC, standard of care.

Context

Key Objective
A clinical researcher is confronted with unique challenges when defining the treatment effect of interest in oncology clinical trials as a result of the diversity of patient journeys. In this article, the relationship between patient journeys and treatment effects is highlighted, and it is illustrated how the newly introduced estimand framework can structure discussions among interdisciplinary teams regarding the design and interpretation of oncology trials.
Knowledge Generated
The estimand framework facilitates a precise definition of the scientific question of interest accounting for different patient journeys. Applying the framework leads to tailored study designs and data analyses.
Relevance
The International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use released a draft addendum of its E9 guideline introducing the concept of an estimand to precisely describe the treatment effect of interest. Understanding the framework and the treatment effect measured in a clinical trial will be important for oncologists so that they can make informed treatment decisions.
There are several steps to confirming the adequacy of the risk-benefit relationship of a new medicine. Addressing it usually requires a sequence of clinical trials, each with its own specific objective. All parties involved in the design of clinical trials need to consider how to define the treatment effect of interest in the study population, along with the relevant end point and data analysis, to adequately address the trial objective in the presence of intercurrent events. Understanding the treatment effect measured in a clinical trial will be important for oncologists so that they can make informed treatment decisions.
The International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use (ICH) aims to align the development of new medicinal products in Europe, Japan, and the United States. The ICH E9 Note for Guidance focuses on the design, conduct, analysis, and reporting of confirmatory clinical trials. However, there is a growing awareness that an imprecise definition of the scientific question of the trial may create a misalignment between its objectives and analysis and confuse its interpretation. In response to this, in 2017, the ICH released a draft addendum1 to its E9 guideline.2 This proposes a structured framework for clinical researchers to better align trial objectives, design, analysis, and interpretation, in view of the diversity of patient journeys and intercurrent events. The addendum introduces the concept of an estimand to precisely describe the treatment effect of interest in terms of the following four attributes: the target population; the end point (or variable) to be obtained for each patient; a specification of how to account for intercurrent events; and a population-level summary for the variable, such as the hazard ratio for PFS.
These four attributes are always interlinked. For example, different strategies for accounting for intercurrent events may result in different end points; more detail is provided in the next section. In addition, each treatment available in the clinical trial should be precisely defined as the (series of) intervention(s) being investigated.
Examples of estimand discussions are surfacing in the oncology literature.3,4 They follow a history of discussions on challenges in the interpretation of clinical benefit end points.5-8 In this article, the relationship between patient journeys and treatment effects is reviewed to illustrate how the estimand framework can structure discussions among interdisciplinary teams about the design and interpretation of oncology trials.

Common Estimand Challenges in Oncology

To illustrate possible challenges when defining estimands for oncology trials, consider the example of the lung cancer trial and the various patient journeys described in the Introduction. The ICH E9 addendum1 describes different strategies for handling intercurrent events to define the treatment effect of interest in a trial. Their use is described in Table 1, focusing on the intercurrent event of starting a new anticancer therapy, as illustrated by the journey of the 64-year-old patient from the Introduction.
Table 1. How the Treatment Effect Definition Drives the Choice of End Point and Handling of an Intercurrent Event in the Estimand Framework for the Lung Cancer Trial Example

Strategies

Treatment policy.

The treatment effect on the end point is of interest regardless of the intercurrent event. For example, estimand 1 (Table 1) targets the effect of treatment on delaying progression or death irrespective of whether a patient receives another therapy. The time until tumor progression or death observed after start of new therapy is still considered to reflect the treatment effect of interest. Choosing the treatment policy strategy could, for example, rely on the consideration that patient journeys that include a change in anticancer treatment reflect the reality of clinical practice, including the possible impact of the investigated treatment on the choice or activity of subsequent therapies. Therefore, it would be relevant to capture this in the definition of the treatment effect of interest for the population. If applied across all intercurrent events, this strategy is in line with the approach that analyzes patients as randomly assigned, disregarding all intercurrent events such as switching treatment (eg, as with the usual intent-to-treat analysis of OS).

Composite.

The occurrence of the intercurrent event provides, by itself, relevant information about the treatment effect of interest. For example, the start of new therapy can be an indicator of lack of efficacy. With the composite strategy, the intercurrent event becomes a component of the end point. For example, estimand 2 (Table 1) considers the effect of treatment on time to progression or death or start of new therapy, all three of which are considered relevant for the assessment of treatment benefit.

Hypothetical.

We are interested in the specific effect of treatment on the end point in the hypothetical scenario in which the intercurrent event does not occur. For example, estimand 3 (Table 1) targets the effect of the studied treatment on PFS as if no alternative therapy existed (ie, excluding its possible impact). In PFS analysis, this strategy has traditionally been attempted by censoring observations at start of new therapy.

Discussion of strategies.

In a clinical trial, different strategies can be used to handle different intercurrent events. This has implications for how we define the end point, as well as how we collect and analyze the data. For example, for estimands 2 and 3, collection of tumor assessments after the start of new therapy is not required, whereas this is necessary for estimand 1. The choice of the estimand will be informed by discussions between all parties involved in the clinical trial, including sponsor and regulators, about the disease setting, the anticipated mode of action of the treatment, and the relevance of intercurrent events.

Example With OS

Table 1 lists the four attributes (population, end point, intercurrent events, and summary) of several estimands for the lung cancer trial, each of which is defined with a PFS end point. It is interesting to further illustrate how the estimand framework can accommodate a patient’s journey when OS (ie, time to all-cause mortality) is the end point of interest. The 64-year-old male patient discussed in the Introduction discontinued treatment as a result of an adverse event and started a new therapy. What is the relevance of this intercurrent event for the treatment effect on OS?
If one assumes that patient journeys observed in the trial after treatment discontinuation reflect real-world clinical practice, the treatment effect of interest can be defined as follows: Does the investigational treatment prolong survival compared with SOC, regardless of treatment discontinuation? This reflects a treatment policy strategy for handling intercurrent events such as the start of new therapy. Hence, the patient’s date of death is relevant for assessing OS, regardless of whether a patient started a new cancer therapy before death. This defines the estimand typically used with OS in oncology trials.
However, it is debatable whether this estimand with OS always yields a clinically meaningful comparison of treatments, in particular if the assumption that subsequent new therapies reflect that clinical practice is violated. For instance, if the protocol allows a crossover, patients from the SOC treatment group may receive the investigational treatment as a new therapy upon disease progression. This does not reflect current clinical practice because this investigational treatment is not yet widely available to patients. Even if it does become available (eg, after regulatory approval based on PFS results), the investigational treatment would likely be used in lieu of SOC, rather than after disease progression on SOC. The comparison of OS between the investigational treatment and the sequence of SOC followed by investigational treatment is, therefore, not assessing a treatment effect of interest. In such a case, the treatment policy strategy to handle the start of new therapy may not adequately address the trial objective. A similar situation arises when patients randomly assigned to SOC have access, upon tumor progression, to other therapies outside the trial with the same mechanism of action (MoA) as the investigational treatment. This was observed in recent immuno-oncology trials, as a result of the multitude of checkpoint inhibitors being investigated.9,10
Using a hypothetical strategy to handle the start of new therapy would target the treatment effect of the investigational treatment on OS, compared with SOC, had no patients from the SOC group subsequently been offered the investigational treatment (or a drug with the same MoA). This may represent a relevant alternative in such situations. In fact, such a hypothetical estimand has been targeted in previous studies with crossover, in addition to the treatment policy estimand.11-13
Knowledge of the evolving treatment landscape, including subsequent lines of therapy, is critical to anticipate possible patient journeys and address estimand challenges with OS. Multidisciplinary discussions at the study design stage with all relevant stakeholders, including regulators, payers, physicians, and patients, are necessary to ensure that an appropriate estimand is chosen.

Example in the Adjuvant Treatment Setting

A 70-year-old woman with a 2.0-cm adenocarcinoma of the breast and no involved lymph nodes comes to her oncologist for treatment. After surgical resection of the tumor, she joins a clinical trial comparing letrozole plus a new drug versus letrozole alone. In her third year after random assignment, she develops a new cancer in the contralateral breast, which is removed surgically. In her seventh year after random assignment, the patient dies from complications after a myocardial infarction.
The scientific question of interest in this adjuvant trial is whether the new treatment prolongs patients’ disease-free survival time. However, a fundamental issue is how to define disease and its recurrence. Should the appearance of a contralateral breast cancer be considered as disease recurrence? And what about another primary cancer or death not related to breast cancer? Such discussions may further refine the scientific question. For example, the researcher may choose between the following three questions to define the objective of this trial: Does the drug delay relapse, secondary malignancy, or death from any cause? Does the drug delay relapse or death from any cause? Does the drug delay relapse or disease-related death?
The choice of scientific question should be informed by a treatment’s MoA and the specific disease setting. For instance, published data suggest that adjuvant hormone therapies for breast cancer may effectively reduce the risk of contralateral breast cancer but increase the risk of other primary malignancies.14-16 Therefore, in this setting, the effect of treatment on the risk of developing these secondary malignancies seems to be relevant. Consequently, the composite strategy could be applied to include these secondary malignancies in the definition of the end point. The researcher would select the first question (Does the drug delay relapse, secondary malignancy, or death from any cause?) as the primary treatment effect of interest for the study.
Such discussions are not new in oncology, and efforts have been undertaken in the adjuvant setting to standardize end points and handle intercurrent events without using the ICH E9 terminology. Hudis et al8(p2127) noted that “standardized definitions of breast cancer clinical trial end points must be adopted to permit the consistent interpretation and analysis of breast cancer clinical trials and to facilitate cross-trial comparisons and meta-analyses.” Recommendations for breast cancer were provided as part of the Definition for the Assessment of Time-to-Event End Points in Cancer Trials initiative, further distinguishing between different types of relapses and secondary malignancies.17 The authors also recognized that handling intercurrent events differently may require new end points. The use of the estimand framework will further facilitate such discussions on study designs and the interpretation of results.

How the Estimand Framework Helped Design A Study of Chimeric Antigen Receptor T-cell Therapy in Lymphoma

A 52-year-old woman visits her oncologist 10 months after receiving rituximab with cyclophosphamide, doxorubicin, vincristine, and prednisone for diffuse large B-cell lymphoma. She has developed a recurrence of her disease, as identified by imaging. Her oncologist reviews different options that could be offered to the patient. The patient could be treated with the SOC of autologous stem-cell transplantation (ASCT). Alternatively, she could receive an investigational personalized therapy, a chimeric antigen receptor T-cell therapy (CAR-T), currently approved to treat patients with diffuse large B-cell lymphoma in a later setting.18,19 The oncologist anticipates the different journeys this patient could encounter on each of the treatment options.
If the patient is to receive SOC, she will initially receive salvage chemotherapy. The best SOC therapy after rituximab plus cyclophosphamide, doxorubicin, vincristine, and prednisone is high-dose chemotherapy followed by ASCT for patients who achieve a deep remission on salvage chemotherapy. ASCT is considered the only possibly curative therapy currently approved for such relapsed disease. If a deep remission is not obtained after two to three cycles of a first salvage chemotherapy (approximately 6 weeks), the patient will receive a new salvage therapy (second salvage), still with the goal to achieve a deep remission and subsequently receive transplantation. A tumor assessment is necessary after first salvage therapy to guide the future course of treatment. Absence of response at this stage would not constitute a final negative outcome of SOC but would determine the next step in that journey. If there was no improvement of disease on two salvage regimens (after approximately 12 weeks into the journey), then SOC would be deemed to have failed. In this situation, the CAR-T therapy could be offered because it is approved in this later setting.
Alternatively, the patient may receive CAR-T immediately, as an experimental option in this earlier setting. There is an expectation that this could also constitute a curative therapy, based on a single infusion. If the patient was to receive CAR-T, her T cells would first be harvested through leukapheresis to enable CAR-T to be manufactured. She might then require chemotherapy to control the disease and prevent rapid progression (bridging chemotherapy) during the approximate 4 to 6 weeks required to manufacture and deliver the CAR-T therapy to the clinic. Within the week before the infusion of the CAR-T cells, blood counts permitting, she would receive lymphodepleting chemotherapy. Although it would be possible that after bridging chemotherapy and before receiving lymphodepleting chemotherapy she would attain remission, there would also be a small chance that her health status would rapidly worsen or that manufacturing would fail, thereby delaying or prohibiting CAR-T administration. However, in general and in contrast with the administration of ASCT in the SOC option, tumor response or progression before CAR-T infusion is not relevant for the administration of CAR-T and, at this stage, would not constitute a final outcome for the patient. If the patient did not receive CAR-T therapy, she might still be a candidate for additional chemotherapy. If there is no improvement of disease after approximately 12 weeks into the journey, then the strategy of CAR-T administration would be deemed to have failed, and other treatment options would need to be considered.
Adopting now the perspective of the researcher who wants to compare both options in a randomized clinical study, a review of the different patient journeys just described raises a number of challenges to identify the adequate research question and the treatment effect of interest for this trial. Those challenges that need to be addressed to design a study enabling a scientifically sound and interpretable comparison are as follows.
First, the therapies to be compared represent two treatment strategies, not simply two different treatments. Both treatment strategies are complex, involving multiple interventions and decision points, and patients may not receive all components of each strategy. The SOC strategy includes one or two lines of salvage therapy, depending on the lack of tumor response to the first line. Patients may or may not receive ASCT, depending on the tumor response after each line of salvage therapy. The CAR-T treatment is a single infusion. However, complexity arises because CAR-T is not available at random assignment. During the manufacturing period, patients may take bridging therapy. Their tumor may progress or respond to the bridging therapy, but the tumor response status before CAR-T infusion is not relevant for the initiation of CAR-T. In addition, manufacturing may fail.
Second, there is an initial period when similar treatment effects are expected from both treatment groups. The bridging therapy given before CAR-T infusion is similar to the SOC salvage therapy. Therefore, the treatment effect of the CAR-T strategy relative to the SOC strategy is not expected to manifest fully during the waiting period.
Finally, CAR-T may be administered after failure of the SOC treatment strategy. For patients receiving the SOC treatment strategy, once two lines of salvage therapy have failed to be effective, patients become eligible to receive the CAR-T therapy, which has already been approved in that setting. This means that the patient outcome observed afterward in the SOC treatment group may not reflect the effect of the SOC strategy alone, but also the effect of this crossover.
The estimand framework helps to address the challenges raised by the complexity of patient journeys on both arms. The researcher will be able to give a more precise description of the treatment effect of interest using the language offered by the estimand framework, starting with a description of the two treatment strategies to be compared.

Treatment Strategies to Be Compared

The study could evaluate the effect of CAR-T against ASCT administration only, focusing on the possibly curative component of each treatment strategy. This would ideally be studied by randomly assigning patients only when they would be ready for both the administration of CAR-T (ie, after successful CAR-T manufacturing) and ASCT (ie, after reaching a sufficient remission from salvage therapy). In this case, patients would be randomly assigned to receive either CAR-T or ASCT, with all necessary preparatory steps in their therapeutic journey being part of a common run-in phase before random assignment.
However, the population for the comparison would then constitute only a subset of the population initially considered eligible to receive the CAR-T or SOC treatment strategies. Randomly assigned patients would have gone through several steps of varying duration since the decision was taken to consider a CAR-T or SOC strategy. It is likely that some eligible patients would not reach the point of random assignment and, therefore, not contribute to the assessment of treatment effect. Patients excluded as a result of insufficient remission could still benefit from CAR-T therapy, but this would not be reflected in the estimated effect. This would not address the following scientific question of interest: What is the relative clinical benefit across the entire patient journey once a CAR-T or SOC treatment strategy is prescribed?
Therefore, it seems of interest to compare the entirety of both treatment strategies instead, including all possible paths as defined in the patient journeys. This also corresponds to clinical practice, where not all patients will receive all components of the treatment.

Population

Aligned with the definition of the treatment strategies described earlier, the population of interest includes all patients eligible to receive either CAR-T or SOC treatment strategies, to be further specified with adequate inclusion and exclusion criteria.

Variable

Typically in this disease setting, the primary variable, or end point, is event-free survival (EFS), representing the time from random assignment to failure of the treatment strategy. Here, failure is defined as first documented disease progression, stable disease, or death as a result of any cause at any time. However, the description of the patient journey clarified that the tumor assessment after the first course of bridging or salvage therapy comes too early to definitely ascertain the outcome of the treatment strategy; disease progression or stable disease is only taken to constitute a failure of a treatment strategy if observed at or after the week 12 assessment. Therefore, a time frame of 12 weeks is used to consider disease progression or stable disease as an event for the EFS end point. The assessment at 6 weeks is only used for therapeutic decision making, and the SOC treatment strategy is further delivered. Thus, it does not enter the EFS calculation because the outcome can still be observed. However, death represents the most informative and definitive outcome of either treatment strategy, which can be observed regardless of when it occurs.

Intercurrent Events

Table 2 lists the intercurrent events affecting the comparison of the two treatment strategies and the estimand strategy used to handle each of them.
Table 2. Intercurrent Events Affecting the Comparison of the Treatment Strategies, With Their Respective Handling Strategy

Manufacturing failure in the CAR-T arm or failing to receive ASCT in the SOC arm.

The patient may not receive the final treatment (CAR-T infusion or ASCT) as a result of the nature of both treatment strategies. Following the treatment policy strategy, such events should be ignored, and patients should still be observed until the final event of interest.

New cancer therapy started before observing EFS event.

If a patient takes an unplanned anticancer therapy before his or her assigned treatment strategy reaches failure, it will be difficult to interpret the outcome as reflecting the effect of that treatment strategy. This intercurrent event can be handled using the hypothetical strategy (ie, to evaluate the effect had patients not been offered such an alternative treatment).

Insufficient remission (stable disease or progressive disease) at week 6.

As discussed earlier, intermediate disease assessments do not determine the final outcome of the treatment strategy. They are ignored, following the treatment policy strategy, as reflected in the variable definition.

Summary Measure

The summary measure often used to quantify the relative treatment effect on EFS in a randomized study is the hazard ratio. This assumes a constant treatment effect over time (proportional hazards assumption). However, in this study, limited or no difference in treatment effect is expected while patients receive similar bridging or salvage therapies. In addition, both CAR-T and ASCT are potentially curative therapies, and hence, a proportion of patients may have a similar outcome when they are cured. These situations violate the proportional hazards assumption, and the hazard ratio may not be an adequate summary measure. Alternative summary measures to quantify treatment effects in the presence of nonproportional hazards are currently being investigated.20-22 Having defined the treatment strategies to be compared, together with the four other attributes of the primary estimand, as summarized in Figure 2, the study design presented in Figure 3 could be proposed.
Fig 2. Estimand components in the phase III chimeric antigen receptor T-cell therapy (CAR-T) case study. ASCT, autologous stem-cell transplantation; DLBCL, diffuse large B-cell lymphoma; PD, progressive disease; SD, stable disease; SOC, standard of care.
Fig 3. The chimeric antigen receptor T-cell therapy (CAR-T) phase III study design. After screening and leukapheresis, patients are randomly assigned to one of the two treatment strategies: CAR-T or standard of care (SOC). In the CAR-T arm, patients may receive bridging chemotherapy during the manufacturing period, before a single CAR-T infusion. Patients are then observed for tumor progression from week 12 onward. In the SOC arm, patients receive up to two salvage chemotherapies, with the aim to receive autologous stem-cell transplantation (ASCT) when disease responds to chemotherapy. If no sufficient response to the first round of chemotherapy is achieved in SOC patients, CAR-T is manufactured in parallel to the patient receiving a new round of chemotherapy. This enables investigators to offer a crossover to CAR-T should lack of response be confirmed at week 12, after the second round of chemotherapy. BIRC, blinded independent review committee; CR, complete response; CT, computed tomography; DLBCL, diffuse large B-cell lymphoma; PD, progressive disease; PET, positron emission tomography; PR, partial response; R, random assignment; SD, stable disease.

Conclusion

Patients with cancer experience different journeys, and researchers need to account for this diversity when defining the treatment effect of interest in a clinical trial. Various events occurring after random assignment may affect the assessment and interpretation of the treatment effect of interest.
The estimand framework provides a structured approach to transparently discuss and account for different patient journeys. Identifying an estimand leads to a precise definition of the scientific question of interest and to tailored study designs and data analyses.23,24
Experience with the estimand framework in clinical research is still limited,25 and its use in oncology is being explored in dedicated regulatory and industry forums and working groups. However, it is fair to say that considerations around diverse patient journeys and their impact on trial end points are not new to the oncology community. The estimand framework is not introducing new complexity in this respect. Rather, it offers a common language to discuss existing complexities with all relevant stakeholders.

Acknowledgment

We thank Aby Buchbinder for his enlightening critical review of the first version of this article. We also thank Bibiana Blatna for her support in formatting this work adequately. We are indebted to Ying Lu, MD, and Jack Lee, MD, for inviting us and gratefully appreciate the opportunity to contribute this review article.

Authors' Disclosures of Potential Conflicts of Interest

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/po/author-center.

Evgeny Degtyarev

Employment: Novartis
Stock and Other Ownership Interests: Novartis

Yiyun Zhang

Employment: Novartis

Kapildeb Sen

Employment: Novartis
Stock and Other Ownership Interests: Novartis

David Lebwohl

Employment: Novartis, Semma Therapeutics
Stock and Other Ownership Interests: Novartis, Semma Therapeutics
Expert Testimony: Novartis

Mouna Akacha

Employment: Novartis
Stock and Other Ownership Interests: Novartis

Lisa V. Hampson

Employment: Novartis, AstraZeneca
Stock and Other Ownership Interests: Novartis

Bjoern Bornkamp

Employment: Novartis
Stock and Other Ownership Interests: Novartis, Alcon
Travel, Accommodations, Expenses: Novartis

Antonella Maniero

Employment: Novartis, Bristol-Myers Squibb
Stock and Other Ownership Interests: Novartis, Bristol-Myers Squibb
Travel, Accommodations, Expenses: Novartis

Frank Bretz

Employment: Novartis
Stock and Other Ownership Interests: Novartis
Travel, Accommodations, Expenses: Novartis

Emmanuel Zuber

Employment: Novartis, Novartis (I)
Stock and Other Ownership Interests: Novartis, Alcon, Air Liquide, Novartis (I), Alcon (I)
Travel, Accommodations, Expenses: Novartis, Novartis (I)
No other potential conflicts of interest were reported.

References

1.
International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use: E9(R1) draft guideline: Estimands and sensitivity analysis in clinical trials. https://www.ich.org/fileadmin/Public_Web_Site/ICH_Products/Guidelines/Efficacy/E9/E9-R1EWG_Step2_Guideline_2017_0616.pdf
2.
International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use: E9 guideline: Statistical principles in clinical trials. https://www.ich.org/fileadmin/Public_Web_Site/ICH_Products/Guidelines/Efficacy/E9/Step4/E9_Guideline.pdf
3.
Rufibach K: Treatment effect quantification for time‐to‐event endpoints: Estimands, analysis strategies, and beyond. Pharm Stat 18:145-165, 2019
4.
Bengoudifa B-R, Weber H, Gathmann I, et al: Estimand framework in AML based on a randomized, placebo-controlled, phase 3 study investigating the effect of adding midostaurin to standard chemotherapy in patients with newly diagnosed AML without a FLT3 mutation. Blood 132:5196, 2018 (abstr)
5.
Fleming TR, Rothmann MD, Lu HL: Issues in using progression-free survival when evaluating oncology products. J Clin Oncol 27:2874-2880, 2009
6.
Korn EL, Freidlin B, Abrams JS: Overall survival as the outcome for randomized clinical trials with effective subsequent therapies. J Clin Oncol 29:2439-2442, 2011
7.
Saad ED, Buyse M: Overall survival: Patient outcome, therapeutic objective, clinical trial end point, or public health measure? J Clin Oncol 30:1750-1754, 2012
8.
Hudis CA, Barlow WE, Costantino JP, et al: Proposal for standardized definitions for efficacy end points in adjuvant breast cancer trials: The STEEP system. J Clin Oncol 25:2127-2132, 2007
9.
Larkin J, Minor D, D’Angelo S, et al: Overall survival in patients with advanced melanoma who received nivolumab versus investigator’s choice chemotherapy in CheckMate 037: A randomized, controlled, open-label phase III trial. J Clin Oncol 36:383-390, 2018
10.
Rittmeyer A, Barlesi F, Waterkamp D, et al: Atezolizumab versus docetaxel in patients with previously treated non-small-cell lung cancer (OAK): A phase 3, open-label, multicentre randomised controlled trial. Lancet 389:255-265, 2017
11.
Reck M, Rodrıǵuez-Abreu D, Robinson AG, et al: Treatment switching–adjusted overall survival (OS) in KEYNOTE-024: First-line pembrolizumab versus chemotherapy in patients with advanced non-small cell lung cancer (NSCLC). J Thorac Oncol 13:112-113, 2018 (suppl 4)
12.
Sternberg CN, Hawkins RE, Wagstaff J, et al: A randomised, double-blind phase III study of pazopanib in patients with advanced and/or metastatic renal cell carcinoma: Final overall survival results and safety update. Eur J Cancer 49:1287-1296, 2013
13.
Motzer RJ, Escudier B, Oudard S, et al: Phase 3 trial of everolimus for metastatic renal cell carcinoma: Final results and analysis of prognostic factors. Cancer 116:4256-4265, 2010
14.
Goss PE, Ingle JN, Martino S, et al: A randomized trial of letrozole in postmenopausal women after five years of tamoxifen therapy for early-stage breast cancer. N Engl J Med 349:1793-1802, 2003
15.
Dong C, Chen L: Second malignancies after breast cancer: The impact of adjuvant therapy. Mol Clin Oncol 2:331-336, 2014
16.
Mellemkjaer L, Friis S, Olsen JH, et al: Risk of second cancer among women with breast cancer. Int J Cancer 118:2285-2292, 2006
17.
Gourgou-Bourgade S, Cameron D, Poortmans P, et al: Guidelines for time-to-event end point definitions in breast cancer trials: Results of the DATECAN initiative (Definition for the Assessment of Time-to-event Endpoints in CANcer trials). Ann Oncol 26:873-879, 2015
18.
Schuster S, Bishop MR, Tam CS, et al: Tisagenlecleucel in adult relapsed or refractory diffuse large B-cell lymphoma. N Engl J Med 380:45-56, 2019
19.
Neelapu SS, Locke FL, Bartlett NL, et al: Axicabtagene ciloleucel CAR T-cell therapy in refractory large B-cell lymphoma. N Engl J Med 377:2531-2544, 2017
20.
Saad ED, Zalcberg JR, Péron J, et al: Understanding and communicating measures of treatment effect on survival: Can we do better? J Natl Cancer Inst 110:232-240, 2018
21.
Uno H, Claggett B, Tian L, et al: Moving beyond the hazard ratio in quantifying the between-group difference in survival analysis. J Clin Oncol 32:2380-2385, 2014
22.
Royston P, Parmar MK: Restricted mean survival time: An alternative to the hazard ratio for the design and analysis of randomized trials with a time-to-event outcome. BMC Med Res Methodol 13:152, 2013
23.
Akacha M, Bretz F, Ohlssen D, et al: Estimands and their role in clinical trials. Stat Biopharm Res 9:268-271, 2017
24.
Permutt T: Effects in adherent subjects. Stat Biopharm Res 10:233-235, 2018
25.
Akacha M, Bretz F, Ruberg S: Estimands in clinical trials: Broadening the perspective. Stat Med 36:5-19, 2017

Information & Authors

Information

Published In

JCO Precision Oncology
Pages: 1 - 10

History

Published online: October 24, 2019

Permissions

Request permissions for this article.

Authors

Affiliations

Evgeny Degtyarev, MS
Novartis Pharma AG, Basel, Switzerland
Yiyun Zhang, PhD
Novartis Pharmaceuticals, East Hanover, NJ
Kapildeb Sen, PhD
Novartis Pharmaceuticals, East Hanover, NJ
David Lebwohl, MD
Novartis Pharmaceuticals, East Hanover, NJ
Mouna Akacha, PhD
Novartis Pharma AG, Basel, Switzerland
Lisa V. Hampson, PhD
Novartis Pharma AG, Basel, Switzerland
Bjoern Bornkamp, PhD
Novartis Pharma AG, Basel, Switzerland
Antonella Maniero, MSc
Novartis Pharmaceuticals, East Hanover, NJ
Frank Bretz, PhD [email protected]
Novartis Pharma AG, Basel, Switzerland
Emmanuel Zuber, PhD
Novartis Pharma AG, Basel, Switzerland

Notes

Frank Bretz, PhD, Novartis Pharma AG, WSJ-27.6.032, Lichtstrasse 35, 4056 Basel, Switzerland; e-mail: [email protected].

Author Contributions

Conception and design: All authors
Administrative support: Emmanuel Zuber
Collection and assembly of data: Evgeny Degtyarev, Kapildeb Sen, Antonella Maniero, Frank Bretz
Data analysis and interpretation: Evgeny Degtyarev, Kapildeb Sen, David Lebwohl, Antonella Maniero, Frank Bretz, Emmanuel Zuber
Manuscript writing: All authors
Final approval of manuscript: All authors
Accountable for all aspects of the work: All authors

Funding Information

Supported by Novartis.

Metrics & Citations

Metrics

Altmetric

Citations

Article Citation

Download Citation

If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download.

For more information or tips please see 'Downloading to a citation manager' in the Help menu.

Format





Download article citation data for:
Evgeny Degtyarev, Yiyun Zhang, Kapildeb Sen, David Lebwohl, Mouna Akacha, Lisa V. Hampson, Bjoern Bornkamp, Antonella Maniero, Frank Bretz, Emmanuel Zuber
JCO Precision Oncology 2019 :3, 1-10

View Options

View options

PDF

View PDF

Get Access

Login options

Check if you have access through your login credentials or your institution to get full access on this article.

Personal login Institutional Login

Purchase Options

Purchase this article to get full access to it.

Purchase this Article

Subscribe

Subscribe to this Journal
Renew Your Subscription
Become a Member

Media

Figures

Other

Tables

Share

Share

Share article link

Share