Pharmaceutical Industry Perspectives on Exposure‐Response Confounding in Large Molecule Therapeutics: Results From an IQ Consortium Survey

ABSTRACT

Exposure‐response (ER) analyses assessing the efficacy of large molecule therapeutics are often susceptible to confounding, where associations between drug exposure and patient disease severity can obscure true ER signals and complicate dose selection. To better understand the pharmaceutical industry perspectives on this challenge, the IQ consortium conducted a comprehensive survey targeting clinical pharmacologists, pharmacometricians, and statisticians. The survey, completed by 125 individuals from 23 pharmaceutical companies, aimed to assess awareness, perceived prevalence, mitigation strategies, and the overall role of ER analyses in the context of confounding. Results revealed strong industry awareness, with 88.0% of respondents acknowledging the relevance of ER confounding. However, uncertainty regarding its pervasiveness persists, particularly in non‐oncology settings (31.2% unsure). A consensus emerged on the value of dose‐ranging study designs as an effective mitigation strategy (81.6% agreement). In contrast, the use of advanced statistical methods for causal inference is inconsistent (45.6% usage), and confidence in their reliability is mixed, with 32.8% of respondents expressing uncertainty and most others rating them only moderate or somewhat reliable. While the majority agreed that confounded ER analyses should be interpreted with caution (79.2%), opinions diverged regarding their value in decision‐making when dose‐ranging data is insufficient. This uncertainty, coupled with a recognized need for additional alignment with health authorities, led to a call for best‐practice guidance (92.8% view as valuable). Overall, the survey findings highlight a need for an industry‐wide common approach and the development of clear frameworks to manage and interpret ER confounding for large molecule therapeutics.

Study Highlights

• What is the current knowledge on the topic?

  • ○
    Exposure‐response analyses for large molecule therapeutics can be confounded when disease severity affects both exposure and clinical outcomes, making causal interpretation difficult and complicating dose selection.
• What question did this study address?

  • ○
    This study surveyed pharmaceutical industry clinical pharmacologists, pharmacometricians, and statisticians to assess awareness, perceived prevalence, impact, mitigation approaches, and decision‐making utility of exposure‐response analyses in the presence of confounding
• What does this study add to our knowledge?

  • ○
    Across 125 respondents from 23 companies, awareness of exposure‐response confounding was high, but views on its prevalence and on the reliability of statistical mitigation methods were mixed. Dose‐ranging studies emerged as the clearest preferred mitigation strategy, and respondents strongly supported development of best‐practice or regulatory guidance.
• How might this change drug discovery, development, and/or therapeutics?

  • ○
    These findings support earlier dose‐ranging and more cautious interpretation of confounded exposure‐response analyses, while highlighting the need for harmonized frameworks to improve dose selection and regulatory communication for large molecule therapeutics.

1. Introduction

Exposure‐response (ER) analysis is a cornerstone of drug discovery and development, providing the quantitative framework necessary to understand the relationship between drug exposure and pharmacological response. By characterizing this link, ER analyses substantiate critical decisions regarding dose justification and support integrated benefit–risk assessments. A central tenet of these analyses is the assumption of causality, that is that differences in drug exposure drive differences in clinical effects. However, this assumption is frequently challenged by confounding factors—particularly those unique to large molecule therapeutics, such as monoclonal antibodies (mAbs), where disease‐specific variables obscure the true relationship between exposure and outcome.

While different forms of ER confounding may affect different treatment modalities, including small and large molecule therapeutics, this survey report focuses mostly on confounding specific to large molecules. These confounders typically arise from the complex interactions between disease biology, patient pathophysiology, and the pharmacokinetics (PK) and pharmacodynamics (PD) of the drug. For large molecules, factors such as baseline disease severity, cancer cachexia, and immunogenicity (specifically anti‐drug antibodies or ADA) can induce disease‐associated clearance changes that fundamentally alter the on‐study drug exposure [1]. While not completely immune to all types of ER confounding, it is important to note that exposure–safety analyses of large molecules may be more robust against such issues, though disease‐related factors such as neutropenia risk in advanced cancer may confound an exposure‐safety analysis.

Recent publications suggest in therapeutic areas like oncology and immunology, patient disease status often acts as a potent confounder of exposure‐efficacy analyses. For instance, patients with advanced disease or rapidly progressing tumors frequently exhibit elevated inflammatory cytokines and hypercatabolism, processes that accelerate the protein turnover and clearance of therapeutic antibodies. Clinical studies of anti‐PD1 antibodies, including nivolumab and pembrolizumab, have demonstrated that patients with higher baseline clearance tend to experience worse overall survival independent of drug exposure, suggesting that the underlying disease severity, rather than insufficient dosing, drives both the lower exposure and the poorer clinical outcome [2, 3].

Mechanistically, these outcomes are thought to be linked to physiological changes in patients with aggressive diseases. Cancer cachexia has been associated with increased proteolytic activity and altered expression of the neonatal Fc receptor (FcRn). Consequently, impaired FcRn‐mediated recycling may be one of the reasons for the shortened half‐life of IgG‐based therapeutics in patients suffering from advanced disease and end‐stage wasting syndromes [4, 5]. Other related mechanisms can occur in autoimmune diseases characterized by systemic immune activation. In rheumatoid arthritis (RA), high baseline levels of immune complexes are thought to competitively saturate FcRn binding, thereby increasing mAb clearance. Studies of infliximab and adalimumab in RA have shown that disease‐related ADA formation further accelerates clearance and reduces serum exposure, driving a correlation of their pharmacokinetic profile with poorer clinical response [6]. Consequently, it becomes difficult to separate the impact of low drug exposure from the impact of severe baseline disease, as efficacy appears limited to those with milder proteinuria. This confounding is underscored by the fact that while simulations suggest a higher dose might offset rapid clearance, there is no clinical evidence that simply increasing the dose improves outcomes for these patients [7].

Hence, accurately interpreting large molecule ER relationships in patients with severe disease presents unique challenges. These patients often possess a poor prognosis independent of treatment, which can create a misleading, non‐causal correlation where low drug exposure and poor outcomes co‐occur—not due to under‐dosing, but because disease severity drives both reduced exposure and diminished efficacy [8]. A clear early example of this phenomenon was observed in HER2‐positive cancer patients with advanced disease, high tumor burden, or systemic inflammation treated with trastuzumab. These patients demonstrate faster trastuzumab clearance associated with shorter overall survival, reflecting disease‐driven physiological alterations rather than insufficient dosing [9, 10].

Without appropriate adjustment for these large molecule confounding factors, ER data may be misinterpreted to suggest that higher doses will improve outcomes. This could, for example, lead to the implementation of dose optimization strategies or post‐marketing requirements that fail to address the underlying drivers of poor prognosis [11]. In inflammatory joint diseases treated with adalimumab, elevated C‐reactive protein (CRP) and ADA levels have also been associated with increased clearance and reduced serum concentrations. In this case, if immune‐related factors are not appropriately accounted for, analyses may incorrectly attribute treatment failure to low drug exposure, when in fact immune activation is driving both the rapid clearance and the disease severity [12].

Recognizing these collective challenges, a comprehensive survey was conducted to gather industry‐wide perspectives on ER confounding for large molecule development. The survey aimed to: (1) evaluate awareness and experience with ER confounding across the pharmaceutical industry; (2) identify areas of consensus and divergence regarding its mechanisms, impact, and mitigation; and (3) gauge interest in harmonized best practices and regulatory guidance. This paper presents the methodology and results of this survey, offering a quantitative snapshot of current industry views.

2. Methodology

The survey was developed and disseminated by the International Consortium for Innovation and Quality in Pharmaceutical Development (IQ) Large Molecule ER Confounding Working Group. The target audience comprised clinical pharmacologists, pharmacometricians (broadly defined), and statisticians employed by pharmaceutical companies represented within the working group. These organizations actively conduct ER analyses to support clinical dose selection for large molecule therapeutics.

The survey instrument consisted of 27 core questions covering several key domains: current landscape and awareness, perceived prevalence and impact of ER confounding, mitigation strategies, the role and interpretation of ER analyses, regulatory alignment, and mechanistic understanding. A variety of response formats were used, including multiple‐choice, check‐all‐that‐apply, rating scale, and open‐ended text responses. The survey questionnaire is presented in Supporting Information S1. It was distributed electronically to IQ member companies. Responses were collected anonymously to encourage honest and candid feedback. Data were collected at both the individual and company level.

To ensure balanced representation and avoid over‐weighting companies with more respondents, company‐level aggregation rules were applied. A response was attributed to a company if it was the most common response among its respondents, or if at least one individual selected the option in “check‐all‐that‐apply” questions (e.g., for Question 24, a single “yes” response resulted in a “yes” for the company). All analyses were descriptive; no inferential statistics were applied. For quantitative questions, response frequencies and percentages were calculated with both the individual responses and companies as the unit of analysis. Missing values were also accounted for in the denominator.

3. Results

3.1. Respondent Characteristics and Representativeness of the Sample

A total of 125 individuals from 23 distinct pharmaceutical companies participated, averaging approximately five respondents per company; although the specific number of respondents per company was not disclosed. The respondents represented a wide range of organizational sizes, with the majority from large pharmaceutical companies. Results presented here are references in Supporting Information S2 (IQ report Q1 and Q2). Specifically, 43.5% of individuals worked at companies with ≥ 50,000 employees, and 38.7% worked at companies with 10,000 to < 30,000 employees. Experience was heavily concentrated in oncology, with 79.2% of respondents reporting experience in solid tumor oncology and 58.4% in hematology. Significant experience was also reported in inflammation/immunology (37.6%) and rare diseases (32.0%).

The close concordance between individual‐ and company‐level distributions across key descriptors, such as company size (43.5% of companies with ≥ 50,000 employees versus 43.5% of respondents from companies with ≥ 50,000 employees) and therapeutic area (e.g., 87.0% of companies versus 79.2% of individuals working in solid tumor oncology), indicates that the individual responses are broadly representative of the participating companies.

3.2. Landscape of ER Analysis and Awareness of Confounding

The application of ER analyses for large molecules is a widespread and frequent practice. Results of this section are provided in Supporting Information S2 (IQ report Q3, Q3a, Q3b, and Q4 response). Over 90% of individual respondents (91.2%) confirmed their organization applied ER analyses for oncology molecules in the past 5 years, with 60.0% confirming the same for non‐oncology molecules. Among those whose organizations used ER analyses, the vast majority reported they are applied “Almost always” (80.3% of individuals, 95.7% of companies). These analyses are most often considered jointly informative and supportive for dose selection (66.4% of individuals from 82.6% pharma companies). Awareness of the potential for ER confounding is also exceptionally high. A large majority of both individuals (88.0%) and companies (95.7%) reported that their organization is aware of the issue.

3.3. Perceptions of Prevalence and Impact

Despite the high awareness of ER confounding, there was no clear consensus on the pervasiveness of confounding (Figure 1). In oncology, individual opinions were distributed across “Sometimes” (28.0%), “In most cases” (24.0%), and “Almost always” (21.6%), with no single dominant view for ER confounding. In non‐oncology, uncertainty was much higher; the most frequent response after “Sometimes” (34.4%) was “Unsure” (31.2%), indicating a possible knowledge gap. Overall, most respondents view confounding as a manageable issue that “Requires Caution” (51.2%) or is “Manageable” with mitigation strategies (32.8%), rather than an insurmountable “Significant Concern” (10.4%).

FIGURE 1.

Perceived Pervasiveness of Exposure‐Response Confounding in (A) Oncology versus (B) Non‐Oncology Settings. Figures display Individual response summaries related to questions 7.4a and 8.4b of Supporting Information S2, including missing values.

Respondents identified certain therapeutic modalities as being more susceptible. IgG‐based monoclonal antibodies, bispecific/trispecific antibodies, and antibody‐drug conjugates (ADCs) received the highest mean susceptibility scores (3.85, 3.85, and 3.79 out of 5, respectively). Conversely, cell therapies, gene therapies, oligonucleotides, and vaccines were rated as less susceptible. Regarding clinical endpoints, Overall Survival (OS) and Progression‐Free Survival (PFS) were identified as most influenced by confounding in oncology (both selected by 68.8% of individuals). However, for both oncology and non‐oncology endpoints, a substantial portion of respondents (39.2% for each) selected “Unsure,” highlighting a gap in specific knowledge (Supporting Information S3, Figure S1).

3.4. Mitigation Strategies and Study Design

There was an overwhelming agreement on the importance of dose‐ranging studies (Figure 2), often incorporating randomization (Supporting Information S3, Figure S2), to address confounding. When asked if dose‐ranging is vital when confounding is suspected, 81.6% of individuals and 95.7% of companies responded “Yes”. Further, 74.4% of individuals from 87% participating pharma companies have identified Phase 2 dose‐ranging trials as the stage where ER analyses are most useful and important.

FIGURE 2.

Consensus on Dose‐Ranging as the Primary Mitigation Strategy (A) and Phase 2 as Stage of the Drug Development Exposure‐efficacy Most Valuable (B). Figures display Individual response summaries related to results 15 in Supporting Information #2 (A) and results 37 (B), including missing values.

Although randomization was the most frequently selected mitigation strategy (36% of individuals), this question was implemented as a single‐choice item rather than a multiple‐choice question, which likely underestimated broader consideration of other strategies. Notably, 27.2% of individuals selected “Unsure,” indicating a lack of general consensus or best practice for mitigation techniques beyond dose‐ranging.

The use of specific statistical or causal inference methods to mitigate confounding was inconsistent. At the individual level, respondents were nearly evenly split, with 45.6% reporting use of such methods and 54.4% reporting they did not. At the company level, usage was more prevalent, with 78.3% of companies having at least one respondent report their use. However, confidence in the reliability of these methods was mixed. The most common response regarding their ability to discern the “ground truth” was “Unsure” (32.8%), followed by “Moderately reliable” (28.0%) and “Somewhat reliable” (27.2%) (Figure 3).

FIGURE 3.

Respondent Confidence in the Reliability of Statistical/Causal Inference Methods. Figures display individual response summaries related to results #34 in Supporting Information S2, including missing values.

3.5. Role and Interpretation of ER Analyses in Confounded Scenarios

Respondents see value in ER analyses to support dose justification even when other more direct evidence may be available. A majority agreed (“Yes” or “Maybe”) that ER analysis has added value even when a dose‐efficacy analysis provides sufficient evidence to support dose selection. However, opinions were divided on relying on confounded ER when dose‐ranging evidence is insufficient. While 44.0% would endorse its integration into the decision‐making process, 34.4% responded “Maybe,” indicating conditional support. In the most difficult scenario—where confounding is deemed impossible to dissociate from a true effect (e.g., in a single‐dose study)—there was no consensus on the usefulness of ER analysis for dose justification, with responses split between “Yes” (28.0%), “No” (35.2%), and “Unsure” (34.4%). Separately, when confounding is likely, 79.2% of respondents indicated that ER analyses should be conducted but interpreted with caution.

3.6. Mechanistic Understanding, Regulatory Alignment, and Need for Guidance

Respondents identified increases in non‐specific clearance (e.g., hypercatabolism) (48.8%) and target‐mediated drug disposition (TMDD) (25.6%) as the most plausible biological mechanisms underlying large molecule ER confounding. Free‐text responses clarified that this non‐specific clearance was related to disease severity, tumor burden, and inflammation‐driven hypercatabolism. From a pharmacometrics standpoint, a majority (50.4%) believe confounding is driven by a combination of altered baseline clearance and time‐dependent PK. This is an important point to understand as many recent publications regarding ER confounding mitigation approaches have focused exclusively on the impact of time‐varying clearance and exposure metric considerations.

Notably, there is considerable uncertainty regarding alignment between industry and health authorities on sources of ER confounding and mitigation approaches. For oncology, 47.2% of individuals perceive alignment, while 22.4% perceive misalignment and 27.2% are unsure. For non‐oncology, this uncertainty dominates, with 50.4% of individuals selecting “Unsure.” This perceived gap fuels a strong desire for clarity. An overwhelming majority of respondents rated the establishment of a best‐practice White Paper or formal regulatory guidance on this important topic as “Very Valuable” (72.8%) or “Somewhat Valuable” (20.0%) (Figure 4).

FIGURE 4.

Perceived Value of Establishing Best Practice or Formal Regulatory Guidance on Large Molecule ER Confounding. Figures display individual response summaries related to results #25 in Supporting Information #2, including missing values.

4. Discussion

A comprehensive understanding of ER confounding is pivotal for the development of large molecule therapeutics, as it fundamentally influences the reliability of dose selection. The presence of ER confounding can obscure or distort a true relationship between drug exposure and clinical outcomes, risking inaccurate interpretations of a drug's benefit–risk profile in the context of dose selection. Such misinterpretations have the potential to precipitate suboptimal dose selection, regulatory setbacks, or even negative patient outcomes. Therefore, effectively identifying and mitigating ER confounding strengthens the validity of ER analyses, enabling more informed clinical and regulatory decision‐making to maximize patient benefit.

This survey, conducted by the IQ Consortium ER Confounding Working Group, provides a comprehensive overview of industry perspectives on ER confounding for large molecule therapeutics. The findings reveal a field that, while broadly aware of the issue, displays considerable diversity regarding its specific mechanisms, potential implications, and management approaches. Four key narratives emerge from the survey data, defining the current landscape.

First, an apparent paradox exists. Despite high awareness of ER confounding, there is substantial uncertainty regarding its real‐world prevalence. Respondents demonstrated limited consensus on where ER confounding is most likely to arise, which clinical settings are most susceptible, and which specific endpoints are impacted. The heterogeneity in perspectives, combined with a high frequency of “Unsure” responses regarding pervasiveness, particularly outside of oncology, suggests that while ER confounding is recognized conceptually, practical experience remains limited. This disconnect between theoretical awareness and application may contribute to an underestimation of risk in specific drug development programs.

Second, the survey highlights a pragmatic industry preference for traditional study design over advanced statistical methods. There is strong consensus that dose‐ranging trials, often incorporating randomization, are the most effective mitigation strategy. This preference for designing around the problem contrasts with significant hesitation regarding causal inference and other advanced statistical techniques, for which confidence remains mixed. Methods such as covariate‐adjusted ER analysis, early‐treatment exposure metrics accounting for time‐varying PK (e.g., first‐cycle exposure), and longitudinal PK/PD modeling are frequently viewed as exploratory rather than definitive solutions [13]. For many practitioners, it appears that these tools are not yet trusted as primary solutions for disentangling a confounded signal, but rather supportive measures to help guide interpretation. Moreover, in scenarios where data is limited, many agree that even advanced statistical techniques may fail to fully resolve confounding. It is often the case that these solutions rely on an unverifiable assumption of no unknown or unmeasured confounding factors. Thus, a potentially important component of ER analyses is gauging confidence that such “latent” confounding variables do not exist and that the correct confounding mechanisms are understood.

Third, while there is broad agreement that confounded ER analyses should be interpreted with caution for dose selection support, consensus fractures when faced with the ambiguity introduced by specific data limitations. Nearly all respondents acknowledged the need for caution; however, opinions diverged on the utility of confounded ER analysis for decision‐making when primary evidence from dose‐ranging data is weak or absent. This lack of alignment reveals a critical gap in industry best practices and further highlights uncertainty on the reliability of confounding mitigation strategies, potentially leading to inconsistent decision‐making across organizations. Nonetheless, the data also suggest some divergent opinions when dose‐efficacy data/analysis are strong enough for dose selection (survey data Q16). Based on these findings, early dose optimization, especially in Phase II, may be critical for characterizing the exposure‐efficacy relationship, identifying an optimal dose that balances efficacy and safety, and improving the likelihood of clinical and regulatory success.

Finally, the survey indicates heterogeneity of opinion regarding industry alignment with health authorities may contribute to this uncertainty. While nearly half of respondents perceive alignment in oncology settings, a substantial combined proportion reported either uncertainty or perceived misalignment with health authorities on the topic. Indeed, misalignment with health authorities could potentially result in superfluous ER analyses in regulatory submissions, ultimately diverting valuable resources and impeding innovation within the pharmacometric community. In non‐oncology settings, uncertainty regarding industry alignment with regulatory expectations was the dominant view. This prevailing uncertainty may either reflect a relative lack of experience in non‐oncology settings for survey respondents or could be a byproduct of the standard practice of assessing multiple dose levels in Phase 2 and Phase 3 non‐oncology studies; ie, since these designs inherently help mitigate confounding, specific regulatory feedback on confounded ER analyses may be less frequently encountered. Nevertheless, this overall ambiguity regarding “best practice” expectations appears to complicate development planning and regulatory strategy for sponsors. This is likely to drive the high demand (92.8%) for formal best‐practice White Paper or regulatory guidance. To our knowledge, existing guidance, such as the FDA's Exposure‐Response Relationships [14] addresses general bias and confounding challenges at a high‐level but does not explicitly outline how ER results should be utilized to inform regulatory and dose decisions when confounding is suspected, particularly in scenarios where data limitations prevent robust statistical correction (e.g., FDA guidance on ER [14]; see also EFPIA MID3 [15] comments calling for greater operational clarity). Specific guidance tackling these issues would likely help direct more efficient use of sponsor and health authority resources and improve development outcomes for patients.

It is widely acknowledged that time‐dependent confounding factors affecting drug accumulation, dose changes, and event onset timing can significantly bias ER analyses, particularly in oncology. Several employees at the U.S. FDA recently published a paper highlighting regulatory expectations for rigorous ER analyses in the context of dose optimization [8]. The authors reinforced the risks of time‐dependent confounding and advocated for the use of multiple exposure metrics, including static exposure measures, to assess robustness. Furthermore, they underscored the importance of interpreting potential causal effects through a framework that integrates ER findings with robust clinical data generation, including dose–response analyses. Their analyses, focused on exposure accumulation and dose modification (interruptions/reductions), have shown that time‐dependent exposure metrics may cause misleading inverse or positive ER trends due to imbalances in the time‐ at‐risk for the accumulation or dose modification events. To mitigate these biases, the authors recommend (1) preferring time‐independent exposure metrics; (2) using model structures that accurately reflect underlying PD relationships; (3) considering trial designs with lower dose levels to better characterize dose–response relationships; (4) assessing consistency between ER and dose–response data; and (5) employing a holistic, evidence‐based approach to support causal inference. While these strategies aim to enhance the accuracy of ER analyses and improve oncology dose optimization within the context of on‐trial dose modifications, they do not explicitly address the unique concerns introduced by large molecule ER confounding factors, where the disease biology itself can inherently drive ER confounding through interactions with clearance pathways. Moving forward, harmonizing approaches, clarifying expectations, and providing a predictable framework for managing large molecule ER confounding are critical next steps to strengthen the scientific basis for dose selection and inform regulatory decision‐making.

5. Limitations

This study has several limitations that should be considered when interpreting the findings. While the examples presented here are illustrative, this survey was not designed to provide an exhaustive mapping of historical cases or a systematic evaluation of mitigation strategies. A more comprehensive synthesis of case studies and corresponding solutions would be a valuable direction for future work to further support practical and rational decision‐making in pharmacometrics.

Further, the survey respondents were from IQ Consortium member companies, which may limit the diversity of perspectives. This limitation arises both from the geographic focus on US‐based organizations and from the exclusion of broader representation across the pharmaceutical industry, particularly smaller biotechnology companies. The respondent pool was disproportionately composed of individuals with oncology expertise, which may explain the greater uncertainty observed in responses to questions outside of the oncology domain. Limitations in the design of two questions (Q19 and Q27) restricted their intended use as “check‐all‐that‐apply” or ranking questions, respectively, although clear primary selections still emerged. Finally, the results represent self‐reported individual perspectives and experiences and do not necessarily reflect the official positions of their respective companies or an industry‐wide consensus.

A related limitation of this survey is that it did not explicitly capture the use of “standard” covariate adjustment combined with population simulation as a causal inference approach. This workflow, while often not framed in causal terminology, corresponds closely to the g‐computation framework [16] and is likely among the most commonly applied causal inference methods in pharmacometrics. As such, our results may underestimate the extent to which causal inference principles are already embedded in routine practice when they are not explicitly labeled as such.

6. Conclusion

The findings of this survey confirm that ER confounding for large molecule therapeutics is a well‐recognized challenge across the pharmaceutical industry. While the existence of confounding is universally recognized, perspectives diverge significantly on its prevalence, impact, and the reliability of statistical management in the absence of dose‐ranging data. This uncertainty, coupled with ongoing demand for formal guidance, underscores the need for a collaborative effort between industry stakeholders, academia researchers, and regulatory agencies to establish and promote harmonized best practices. Developing a more structured framework for risk assessment, mitigation strategies, and interpretation will be essential toward ensuring understanding of ER confounding and to reliably characterize “true” ER relationships for efficient clinical development of safe and effective large molecule therapeutics.

Author Contributions

E.S., M.L., S.A.‐O., E.W., D.S., Z.L., T.L., S.S., L.A., W.G., X.G., H.G., D.M., and D.C.T. wrote the manuscript; D.C.T., M.L., E.S., H.G., D.S., W.G., X.G., D.M., and E.W. designed the research; D.C.T., M.L., E.S., H.G., D.S., W.G., X.G., D.M., and E.W. performed the research; D.C.T., M.L., E.S., T.L., and S.S. analyzed the data.

Funding

The authors have nothing to report.

Conflicts of Interest

All authors were employees of their respective institutions at the time of preparing this manuscript and may own company stock. Any expert opinions in this manuscript are those of the authors and may not necessarily reflect the view of their respective institutions. This manuscript was developed with the support of the International Consortium for Innovation and Quality in Pharmaceutical Development (IQ, www.iqconsortium.org).

Supporting information

Data S1: Supporting Information.