A Model-Based Meta-Analysis of Pembrolizumab Effects on Patient-Reported Quality of Life: Advancing Patient-Centered Oncology Drug Development.

Study Highlights What is the current knowledge on the topic? ○ Patient‐reported outcomes (PROs), particularly patient‐reported quality of life (QoL) measures, are essential for patient‐centered oncology care. ○ However, PROs data are often underutilized in clinical decision‐making due to their subjective nature, high variability, and limited alignment with objective clinical endpoints. ○ Pembrolizumab, a PD‐1 immune checkpoint inhibitor, has demonstrated significant survival benefits across multiple cancer types. ○ Yet, many clinical trials have reported no significant improvements in QoL with pembrolizumab treatment. What question did this study address? ○ This study addressed the questions of whether pembrolizumab improves patient‐reported QoL and whether the population nonlinear mixed‐effects (NLME) model can uncover treatment benefits on QoL that may be missed by traditional statistical approaches. What this study adds to our knowledge? ○ This model‐based meta‐analysis study shows that the NLME modeling approach effectively characterizes treatment effects of pembrolizumab on patient‐reported QoL by leveraging longitudinal data. ○ It also highlights the limitations of conventional analyses and the urgent need for standardized PRO instruments and reporting/analysis practices. How this might change clinical pharmacology and therapeutics? ○ This study supports integrating NLME modeling into PROs data analysis to enhance the detection of treatment effects, thereby promoting patient‐centered decision making in oncology drug development and informing future regulatory evaluations.

Introduction

As oncology drug development continues to evolve, there is a growing emphasis on patient‐centered care. Nowadays, treatment success is measured not only by survival outcomes but also by improvements in patient quality of life (QoL) [ 1 ]. Patient‐reported outcomes (PROs), particularly those assessing QoL, are recognized as essential endpoints in cancer clinical trials [ 2 , 3 ]. These outcomes provide a direct, patient‐centered view of treatment benefits, symptom burden, and overall well‐being [ 4 ]. Despite their clinical importance, PROs are often underutilized in regulatory and clinical decision‐making due to the high variability and missingness, subjective reporting, lack of standardization instruments, and inconsistent alignment with clinical survival endpoints [ 5 , 6 , 7 ].

Pembrolizumab, a PD‐1 immune checkpoint inhibitor approved for over 20 indications, has shown significant survival benefits across many cancers [ 8 ]. Yet, its impact on patient‐reported QoL remains unclear. Many trials have reported no significant differences in QoL between pembrolizumab and control groups, despite clear survival advantages [ 9 , 10 , 11 ]. This inconsistency may be explained by the high variability in patient‐reported data, which limits the power of traditional hypothesis testing [ 12 ]. Additionally, most studies only assessed changes from baseline to endpoint, overlooking the importance of longitudinal trajectories of QoL. Prior research suggests that QoL trajectories are more strongly associated with survival than single time‐point assessments [ 13 , 14 ].

To better understand the effects of pembrolizumab on patient‐reported QoL, we conducted a model‐based meta‐analysis (MBMA) of published clinical trials across multiple cancer types. We applied population nonlinear mixed‐effect (NLME) modeling to analyze longitudinal patient‐reported QoL data for pembrolizumab, accounting for both between‐study and between‐arm variability. This approach enables a more precise assessment of pembrolizumab's benefits on QoL. Our analysis also highlights the potential of model‐based methods to improve the evaluation of PROs in oncology drug development. By providing a scalable and quantitative framework for analyzing PROs data, this study supports the broader adoption of patient‐centered endpoints in drug evaluation, regulatory review, and clinical decision‐making.

Methods

Literature Search

We conducted a PubMed search ( http://www.ncbi.nlm.nih.gov/pubmed/ ) using the query: ([Pembrolizumab] AND [melanoma]) AND (patient‐reported OR quality of life OR PRO‐CTCAE), with the filter “Clinical Trial” to include only clinical studies reporting PROs for pembrolizumab in melanoma. The term “melanoma” was then replaced with other cancer types, including non‐small‐cell lung cancer (NSCLC), triple‐negative breast cancer (TNBC), head and neck squamous cell carcinoma (HNSCC), urothelial cancer, and endometrial/colorectal cancer (CRC), to identify studies across multiple indications. Studies without longitudinal patient‐reported outcome (PRO) data were excluded. We further restricted the final database to studies reporting longitudinal EORTC QLQ‐C30 Global Health Status/Quality of Life (GHS/QoL) scores, as this is the most widely used PRO instrument in oncology clinical trials [ 15 ]. The GHS/QoL scores (referred to as QoL scores in this study) are derived from two items in the EORTC QLQ‐C30 questionnaire: How would you rate your overall health during the past week? How would you rate your overall quality of life during the past week?

Each item is rated on a 7‐point scale ranging from 1 (very poor) to 7 (excellent). The average of the two scores is then linearly transformed according to EORTC scoring guidelines [ 16 ]. The resulting longitudinal QoL scores range from 0 to 100, with higher scores indicating better overall health status and QoL.

Data Extraction

This MBMA study categorized treatment arms into two groups: pembrolizumab arms and control arms. Pembrolizumab arms included all treatment arms that administered pembrolizumab, either as monotherapy or in combination with other agents. Control arms referred to the comparator treatments from the original clinical trials, typically consisting of placebo plus standard of care. These regimens varied by indication and included chemotherapy, targeted therapy, or placebo alone, depending on the specific trial design.

Longitudinal QoL scores from either pembrolizumab or the control arm were extracted from published figures or tables using WebPlotDigitizer ( https://automeris.io/ ). All included data were required to have at least three QoL measurements per arm; studies reporting only baseline and endpoint QoL scores were ineligible in our analyses. Variability measures associated with the QoL scores were digitized and recorded when available. Additional trial‐level information including study design, sample size, patient age, gender, tumor type, disease stage, and Eastern Cooperative Oncology Group (ECOG) performance status was also collected to construct the MBMA database. All data were collected as aggregate‐level summaries by study arm.

Longitudinal QoL Trajectory Structural Model Development

To characterize the longitudinal trajectories of QoL scores, we developed a semi‐mechanistic population model using data from the MBMA database. In this model, treatment‐related side effects were captured as a decline in QoL scores using an asymptotic exponential function, characterized by two parameters: E max (maximum toxicity) and K p (rate constant for toxicity onset). The treatment‐related improvement in overall well‐being was modeled using a linear function with a slope parameter (SLP). The QoL trajectory structural model is presented in Equation ( 1 ). (1) E t = E 0 − E max × 1 − e − Kp × t + SLP × t where E ( t ) is the QoL score change over time, E 0 is the QoL score at baseline, SLP is the QoL improvement rate, E max is the maximal toxicity, and K p is the toxicity offsite rate (1/week), t is the time in weeks.

Random Effects Model Development

To ensure that model‐predicted QoL scores stayed within the 0 to 100 range, a logit transformation was applied to the observed scores as the model input. The transformed value is defined in Equation ( 2 ): (2) trans y ijk = N ijk × log it y ijk / 100 where y ijk is the observed QoL scores in study i , at the time point j , and treatment arm k . The term trans y ijk represents the logit‐transformed y ijk used as the dependent variable (DV) in the model. N ijk is the number of individuals corresponding to that observation.

A constant residual error weighted by sample size N ijk was used to describe residual variability in Equation ( 3 ). (3) trans y ijk = log it E ijk + ϵ ijk , ϵ ijk ~ N 0 σ 2 N ijk where E ijk is the model‐predicted transformed QoL, and ϵ ijk is the constant residual error following a normal distribution with mean 0 and variance σ 2 scaled by sample size N ijk .

To capture heterogeneity across studies and treatment arms, between‐study variability (BSV) and between‐treatment arm variability (BTAV) were incorporated into model parameters. For parameters bounded between 0 and 100, such as baseline QoL score E 0 , a logit transformation was applied to map them to the continuous parameter space when estimating random effects, as shown in Equation ( 4 ): (4) log it E 0 ik = log it E 0 pop + η E 0 i 0 + η E 0 ik 1 where E 0 ik is the baseline QoL scores for study i and treatment arm k . The E 0 pop is the population parameter for E 0 . The term η E 0 i 0 and η E 0 ik 1 represents the BSV and BTAV following a normal distribution with mean 0 and standard deviation Ω BSVE 0 and Ω BTAE 0 , respectively.

Parameter estimation was performed using a nonlinear mixed‐effects (NLME) modeling approach implemented via the Stochastic Approximation Expectation–Maximization (SAEM) algorithm with a linear method.

Covariate Model Development

The effects of pembrolizumab or control arm on longitudinal QoL trajectories were modeled as covariates using an exponential function (Equation 5 ): (5) P ik = θ P × e Cov trt , Cov trt = 0 if treatment = Pembrolizumab θ P _ control if treatment = Control where P ik is the individual value for parameter P in study i arm k , and θ P is the typical value for parameter P , θ P _control is the estimable parameter for the effect of the control arm relative to pembrolizumab on parameter P .

In addition to treatment effects, other baseline covariates were evaluated based on exploratory analysis and clinical relevance. These covariates included age, proportion of male patients, proportion with stage IV disease, proportion with ECOG performance status of 0, and tumor type. Covariates were explored for their potential influence on key model parameters E 0 , E max , and SLP.

A full fixed‐effects modeling approach was used, in which all relevant covariates were added to the base model simultaneously. Covariate correlations were assessed, and when strong correlations were present, the more clinically relevant or statistically significant covariate was retained. Categorical covariate effects were modeled as shown in Equation ( 5 ). Continuous covariates were modeled using an exponential function (Equation 6 ). For example, the effect of Age on parameter P was described as: (6) P i = θ P × e θ Age × Age Age _ Ref where θ Age is the estimable parameter for the effect of age, and Age_Ref is the reference value for age.

Missing values for covariates were imputed with the population median (for continuous covariates) or mode (for categorical covariates) during the covariate analyses.

Assessment of Model Performance

Model performance was evaluated based on changes in the objective function value (OFV), diagnostic plots, visual predictive check (VPC), and assessments of parameter precision and shrinkage. Diagnostic plots included observations versus population predictions or individual predictions, individual weighted residuals versus time or predictions, and individual predictions over time overlaid with observations [ 17 ]. VPCs were conducted using 1000 simulations to compare observed data with model‐predicted intervals, assessing the model's ability to capture both central trends and variability in the data [ 18 ]. Parameter precision was evaluated using relative standard error (RSE), with values below 30% indicating reliable and stable estimates. Shrinkage, calculated for parameters with inter‐individual variability (IIV), reflects how much individual estimates are pulled toward the population mean. Shrinkage below 30% suggests adequate capture of BSV and BTAV, while higher shrinkage may indicate unreliable individual predictions even when overall diagnostics appear acceptable [ 19 , 20 ].

Model Selection

Several structural models were examined during model development, including linear and exponential forms of QoL improvement, with or without treatment‐related toxicity. Different configurations of BSV and BTAV were tested across model parameters. Each candidate model was assessed using diagnostic plots, VPC, and parameter estimates. The final model was selected based on better goodness‐of‐fit, improved VPC performance, acceptable parameter precision and shrinkage, and a lower OFV.

Model Validation

To assess model stability and parameter uncertainty, nonparametric bootstrap validation was conducted with 500 replicates. In each replicate, a new dataset was generated by random sampling with replacement from the original dataset at the study arm level, preserving the longitudinal structure within each arm. The model was re‐estimated for each bootstrap replicate using the same structural and statistical model. The distribution of parameter estimates across replicates was used to construct parameter 10th/90th percentiles, evaluate estimation robustness, and assess the precision of fixed‐effect and random‐effect parameters.

Mode Simulation

The longitudinal impact of pembrolizumab on QoL trajectories was evaluated using model‐based simulations and compared with the control arm. A total of 500 simulations were conducted, incorporating BSV and BTAV to capture individual heterogeneity.

Statistical Analysis

The effects of treatment type (pembrolizumab versus control) on QoL trajectory parameters will be compared using Wilcoxon tests. A two‐sided p < 0.05 will be considered statistically significant.

Software and Code Availability

MBMA model was developed using Monolix 2024R1. The model simulations were performed using Simulx 2024R1. Both Monolix and Simulx can be downloaded at https://lixoft.com/products/ . Plots and survival analyses were performed using R 4.4.1 and RStudio Version 2022.07.1+554. The figures were reorganized in GraphPad Prism (Version 10.4.1). The model codes are provided in the Supplementary Codes .

Results

MBMA Database

Initial literature screening was completed by March 16, 2025, identifying 302 studies across 7 tumor types. Studies were included if they reported longitudinal QoL scores, resulting in 20 studies selected for final inclusion in the MBMA database. The inclusion and exclusion processes are illustrated in Figure S1 . Study information, along with the number of available QoL data points and sample sizes, is summarized in Table 1 . Of the studies comparing pembrolizumab to control, 8 reported improved QoL with pembrolizumab, while 8 found no significant or clinically meaningful difference between groups. In total, 410 QoL data points were collected, with 228 from pembrolizumab arms and 182 from control arms. The aggregated longitudinal QoL trajectories are shown in Figure 1A . To assess potential bias, we compared the QoL effect size between the pembrolizumab and control arms against data variability using a funnel plot (Figure 1B ). Most data points fell within the 95% confidence interval, indicating no apparent publication or selection bias in the dataset.

MBMA Model Analysis

The model describing longitudinal QoL trajectories was developed using the MBMA database and demonstrated a good fit to the data. BTAV and BSV were incorporated into the baseline QoL score ( E 0 ), while only BSV was included for the maximal toxicity effect ( E max ). Additionally, BTAV was added to the QoL improvement rate (SLP) without logit transformation. Final model diagnostic plots showed no apparent bias in either the pembrolizumab or control arms (Figure 2A ). The VPC plot indicated that the model's prediction intervals adequately captured the observed data, supporting strong model performance (Figure 2B ). Figure 3 presents observed QoL scores over time alongside model‐predicted trajectories for each individual study. Despite the high variability in the data, the model successfully captured the overall trends and variability in QoL trajectories.

In the final MBMA model, treatment effect was identified as a significant covariate on maximal toxicity ( E max ) and QoL improvement rate (SLP). No additional significant covariates were identified during model development. Parameter estimates from the final model, along with bootstrap results, are presented in Table S1 . All parameter estimates had reasonable RSE, indicating precise and robust estimation. Shrinkage values were all below 30%, suggesting good characterization of between‐study and between‐arm variability.

MBMA Reveals Pembrolizumab Longitudinal QoL Benefits

Using our developed MBMA model, we simulated longitudinal QoL trajectories for patients receiving pembrolizumab versus control (Figure 4A ). Patients treated with pembrolizumab experienced greater initial toxicity but demonstrated sustained long‐term improvements in QoL scores.

To compare longitudinal QoL trajectories between pembrolizumab and control arms, we selected two key MBMA model parameters: the QoL improvement rate ( SLP ) and maximal toxicity ( E max ). Pembrolizumab was associated with a significantly higher QoL improvement rate (SLP, p < 0.0001) and significantly greater initial toxicity ( E max , p = 0.005), compared to the control arm (Figure 4B ). These findings are consistent with prior clinical observations that immune checkpoint inhibitors, such as pembrolizumab, may induce early toxicities (e.g., fatigue, diarrhea, rash) that negatively impact QoL, while offering longer‐term clinical benefits such as prolonged survival and delayed disease progression [ 38 , 39 ].

We further conducted a subgroup analysis focusing on studies that originally reported no between‐group differences in QoL (Table 1 ). Even within this subset, the model‐estimated QoL improvement rate (SLP) remained significantly higher in the pembrolizumab arm compared to control ( p < 0.05), and a borderline significantly higher toxicity ( E max ) was observed ( p = 0.052) in pembrolizumab (Figure 4C ). These results highlight the strength of the NLME modeling approach in accounting for variability in PRO data and capturing longitudinal treatment effects that may be missed by conventional statistical comparisons. Incorporating the full trajectory of longitudinal PRO data, rather than relying solely on baseline‐to‐endpoint comparisons, may offer a more sensitive and informative assessment of treatment impact on patient QoL.

Discussion

As oncology drug development continues to shift from a disease‐centered to a patient‐centered paradigm, it is essential to ensure that therapies not only prolong patient survival but also improve patients' QoL [ 13 , 40 ]. In this study, we conducted an MBMA study to evaluate patient‐reported QoL outcomes associated with pembrolizumab across multiple cancer types. To address the substantial variability commonly seen in patient‐reported QoL data, which often limits the effectiveness of traditional statistical methods, we applied an NLME modeling approach. This method allowed us to incorporate the longitudinal trajectory of QoL data and was found to more accurately capture treatment benefits compared to the statistical approaches used in the original publications. Our analysis showed that although pembrolizumab was associated with greater initial toxicity compared to control, it provided meaningful long‐term improvements in QoL across different cancer populations. These findings offer a strong quantitative foundation to guide more patient‐focused drug development and inform future regulatory decisions.

MBMA offers a robust quantitative framework for integrating data across multiple clinical trials while accounting for between‐study heterogeneity [ 41 ]. In our MBMA study, an initial literature search identified over 300 publications reporting pembrolizumab efficacy data. However, only 6.6% ( N = 20) of these studies included valid longitudinal patient‐reported EORTC QLQ‐C30 QoL data. Most publications were excluded due to the absence of PROs or the use of non‐standardized instruments to assess QoL. Although several guidelines have emphasized the importance of PROs [ 42 , 43 ] and the FDA has recommended their inclusion in early‐phase oncology trials to support dose optimization and patient‐centered treatment [ 44 ], our findings reveal a persistent lack of standardized PRO instruments and data reporting. Moreover, there is a lack of quantitatively derived thresholds for clinically meaningful differences in QoL scores, leading to variability across studies [ 45 ]. Current definitions often rely on empirical or context‐specific criteria, which can differ by cancer type and patient population. These gaps highlight the urgent need to improve and standardize PRO collection, reporting practices, and analyses in future clinical trials to better inform decisions made by patients, clinicians, and regulatory stakeholders.

The PRO data from Phase III clinical trials are rarely incorporated into clinical decision‐making, often due to the lack of significant between‐group differences—despite clear survival benefits. A review of drug approvals between 2011 and 2017 found that only 10 out of 71 (14%) showed PRO benefits at the time of approval [ 46 ]. This has limited the use of PRO data in regulatory and clinical decision‐making. In this study, we used a population NLME modeling approach to analyze longitudinal QoL trajectories and characterize treatment effects on PRO endpoints. Compared to traditional hypothesis tests, the NLME approach more effectively detected treatment‐related differences by incorporating the full trajectory of longitudinal data and quantifying interindividual variability. Our findings demonstrate the value of applying population modeling to PRO analysis and highlight its potential to enhance the integration of PRO into drug development and decision‐making.

A key modeling assumption in this study was the separation of treatment‐related toxicity and drug efficacy in their respective impacts on QoL. Toxicity was modeled as a negative effect decreasing QoL over time with E max and K p parameters, while efficacy was modeled as a positive effect reflecting symptom relief or delayed disease progression with the SLP parameter. Although some degree of interaction or collinearity between these components may exist, we chose this structure to enable clinically interpretable parameters that can inform future patient‐centered dose optimization strategies. While E max and SLP were estimated separately to reflect early toxicity and treatment‐related QoL improvement, we caution against interpreting their individual magnitudes in isolation. Due to their opposing effects, the net impact of treatment on QoL is best assessed through the overall predicted trajectory. In our model, the toxicity effect stabilizes over time, while the treatment‐related improvement accumulates linearly, aligning with the clinical understanding of transient immune‐related adverse events and sustained therapeutic benefits.

Our study has limitations. First, the dataset lacked sufficient power to compare pembrolizumab's QoL benefits across different tumor types. Future research is needed to determine whether pembrolizumab's toxicity profile and QoL effects differ by cancer type. Another limitation is that our MBMA model was developed on arm‐level aggregate data, whereas the statistical analysis on QoL data in the original clinical studies used individual‐participant level data. This creates a gap that limits the direct comparison between pharmacometrics and traditional statistical methods. Further analyses using individual participant‐level data are needed to validate the utility of population modeling in PRO data analysis. Lastly, potential bias could be introduced by informative dropout in QoL data in the analysis. In the original clinical trials, QoL assessments after treatment discontinuation (often due to progression or toxicity) were typically missing and not imputed. As a result, arm‐level QoL trajectories may overrepresent patients with more favorable outcomes, particularly at later time points. This may lead to an overestimation of treatment‐related QoL benefit. Future work should consider approaches such as joint modeling or pattern‐mixture models to better account for dropout and its impact on longitudinal QoL interpretation.

Conclusion

In conclusion, we conducted an MBMA to evaluate the impact of pembrolizumab on patient‐reported QoL across multiple cancer types. By applying an NLME modeling approach, we were able to characterize the longitudinal QoL trajectories and quantify treatment effects that might be confounded by the data variability when using traditional statistical hypothesis tests. Our findings support the use of advanced modeling techniques to improve the interpretation of PRO data and highlight the importance of incorporating longitudinal QoL measures into clinical trial evaluation. Ultimately, this work contributes to a growing body of evidence promoting more patient‐centered approaches in oncology drug development and regulatory decision‐making.

Author Contributions

J.Z. and L.I.W. wrote the manuscript and designed the research. Y.Z. and Y.S. collected the data. S.R. performed analyses.

Conflicts of Interest

L.I.W. was a consultant for Celgene/Bristol Myers Squibb as a member of the Scientific Steering Committee for the Connect Multiple Myeloma patient registry. This consultant work is unrelated to this manuscript. All other authors declared no conflicts of interest for this work. As an Editor‐in‐Training for CPT: Pharmacometrics & Systems Pharmacology, Jiawei Zhou was not involved in the review or decision process for this paper.

Supporting information

Data S1: psp470106‐sup‐0001‐DataS1.pdf.

Acknowledgments

We acknowledge the longitudinal model‐based meta‐analysis tutorial provided by Lixoft ( https://monolixsuite.slp‐software.com/tutorials/2024R1/case‐study‐longitudinal‐model‐based‐meta‐analysis‐ ). This work is funded by the University of North Carolina at Chapel Hill. Artificial intelligence tools (ChatGPT, OpenAI) were used to assist with language editing and improving readability during manuscript preparation.

Data Availability Statement

The data was accessed through published literature. The organized MBMA database is available from the corresponding author upon reasonable request.

References

1 Prioritizing Patient‐Centered Care in a World of Increasingly Advanced Technologies and Disconnected Care ( Seminars in Radiation Oncology , 2024 ). 10.1016/j.semradonc.2024.07.001 PMC12448779 39271280

2 E. Basch , “ Patient‐Reported Outcomes‐Harnessing Patients' Voices to Improve Clinical Care ,” New England Journal of Medicine 376 ( 2017 ): 105 – 108 . 28076708 10.1056/NEJMp1611252

3 S. S. Chang and B. Movsas , How Vital Are Patient‐Reported Outcomes , vol. 114 ( Oxford University Press , 2022 ), 347 – 348 . 10.1093/jnci/djab178 PMC8902334 34508599

4 P. R. Deshpande , S. Rajan , B. L. Sudeepthi , and C. A. Nazir , “ Patient‐Reported Outcomes: A New Era in Clinical Research ,” Perspectives in Clinical Research 2 ( 2011 ): 137 – 144 . 22145124 10.4103/2229-3485.86879 PMC3227331

5 A. Agarwal , T. Pain , J. F. Levesque , et al., “ Patient‐Reported Outcome Measures (PROMs) to Guide Clinical Care: Recommendations and Challenges ,” Medical Journal of Australia 216 ( 2022 ): 9 – 11 . 34897693 10.5694/mja2.51355 PMC9299767

6 M. Fleischmann and B. Vaughan , “ The Challenges and Opportunities of Using Patient Reported Outcome Measures (PROMs) in Clinical Practice ,” International Journal of Osteopathic Medicine 28 ( 2018 ): 56 – 61 .

7 H. Nguyen , P. Butow , H. Dhillon , and P. Sundaresan , “ A Review of the Barriers to Using Patient‐Reported Outcomes (PROs) and Patient‐Reported Outcome Measures (PROMs) in Routine Cancer Care ,” Journal of Medical Radiation Sciences 68 ( 2021 ): 186 – 195 . 32815314 10.1002/jmrs.421 PMC8168064

8 V. Silva and C. Matos , “ Recent Updates in the Therapeutic Uses of Pembrolizumab: A Brief Narrative Review ,” Clinical and Translational Oncology 26 ( 2024 ): 2431 – 2443 . 38658461 10.1007/s12094-024-03491-8

9 M. A. Khattak , J. J. Luke , G. V. Long , et al., “ Adjuvant Pembrolizumab Versus Placebo in Resected High‐Risk Stage II Melanoma: Health‐Related Quality of Life From the Randomized Phase 3 KEYNOTE‐716 Study ,” European Journal of Cancer 176 ( 2022 ): 207 – 217 . 36202690 10.1016/j.ejca.2022.08.004

10 D. Rischin , K. J. Harrington , R. Greil , et al., “ Pembrolizumab Alone or With Chemotherapy for Recurrent or Metastatic Head and Neck Squamous Cell Carcinoma: Health‐Related Quality‐Of‐Life Results From KEYNOTE‐048 ,” Oral Oncology 128 ( 2022 ): 105815 . 35381576 10.1016/j.oraloncology.2022.105815

11 J. D. Spicer , M. C. Garassino , H. Wakelee , et al., “ Neoadjuvant Pembrolizumab Plus Chemotherapy Followed by Adjuvant Pembrolizumab Compared With Neoadjuvant Chemotherapy Alone in Patients With Early‐Stage Non‐Small‐Cell Lung Cancer (KEYNOTE‐671): A Randomised, Double‐Blind, Placebo‐Controlled, Phase 3 Trial ,” Lancet 404 ( 2024 ): 1240 – 1252 . 39288781 10.1016/S0140-6736(24)01756-2 PMC11512588

12 J. Zhou , B. Muluneh , Q. Li , and J. H. Hughes , “ Revolutionizing Patient‐Reported Outcomes Analysis for Oncology Drug Development Using Population Models ,” Clinical Cancer Research 31 ( 2025 ): 1580 – 1586 . 40013954 10.1158/1078-0432.CCR-24-4073 PMC12010960

13 A. S. Bhatt , M. B. Schabath , A. I. Hoogland , H. S. Jim , and R. Brady‐Nicholls , “ Patient‐Reported Outcomes as Interradiographic Predictors of Response in Non–Small Cell Lung Cancer ,” Clinical Cancer Research 29 ( 2023 ): 3142 – 3150 . 37233986 10.1158/1078-0432.CCR-23-0396 PMC10425729

14 J. Zhou , B. Muluneh , Z. Wang , H. Yao , and J. H. Hughes , “ Leveraging Longitudinal Patient‐Reported Outcome Trajectories to Predict Survival in Non‐Small Cell Lung Cancer ,” Clinical Cancer Research 31 ( 2025 ): 2685 – 2694 . 40272273 10.1158/1078-0432.CCR-25-0292

15 P. Fayers , A. Bottomley , and Group, E.Q.o.L ., “ Quality of Life Research Within the EORTC—The EORTC QLQ‐C30 ,” European Journal of Cancer 38 ( 2002 ): 125 – 133 . 10.1016/s0959-8049(01)00448-8 11858978

16 P. Fayers , N. K. Aaronson , K. Bjordal , M. Grønvold , D. Curran , and A. Bottomley , “EORTC QLQ‐C30 Scoring Manual (European Organisation for Research and Treatment of Cancer: 2001)” .

17 T. Nguyen , M. S. Mouksassi , N. Holford , et al., “ Model Evaluation of Continuous Data Pharmacometric Models: Metrics and Graphics ,” CPT: Pharmacometrics & Systems Pharmacology 6 ( 2017 ): 87 – 109 . 27884052 10.1002/psp4.12161 PMC5321813

18 R. Upton and D. Mould , “ Basic Concepts in Population Modeling, Simulation, and Model‐Based Drug Development: Part 3—Introduction to Pharmacodynamic Modeling Methods ,” CPT: Pharmacometrics & Systems Pharmacology 3 ( 2014 ): 1 – 16 . 10.1038/psp.2013.71 PMC3917320 24384783

19 D. R. Mould and R. N. Upton , “ Basic Concepts in Population Modeling, Simulation, and Model‐Based Drug Development—Part 2: Introduction to Pharmacokinetic Modeling Methods ,” CPT: Pharmacometrics & Systems Pharmacology 2 ( 2013 ): 1 – 14 . 10.1038/psp.2013.14 PMC3636497 23887688

20 D. R. Mould and R. Upton , “ Basic Concepts in Population Modeling, Simulation, and Model‐Based Drug Development ,” CPT: Pharmacometrics & Systems Pharmacology 1 ( 2012 ): 1 – 14 . 10.1038/psp.2012.4 PMC3606044 23835886

21 A. Bottomley , C. Coens , J. Mierzynska , et al., “ Adjuvant Pembrolizumab Versus Placebo in Resected Stage III Melanoma (EORTC 1325‐MG/KEYNOTE‐054): Health‐Related Quality‐Of‐Life Results From a Double‐Blind, Randomised, Controlled, Phase 3 Trial ,” Lancet Oncology 22 ( 2021 ): 655 – 664 . 33857414 10.1016/S1470-2045(21)00081-4

22 E. Bührer , M. Kicinski , M. Mandala , et al., “ Adjuvant Pembrolizumab Versus Placebo in Resected Stage III Melanoma (EORTC 1325‐MG/KEYNOTE‐054): Long‐Term, Health‐Related Quality‐Of‐Life Results From a Double‐Blind, Randomised, Controlled, Phase 3 Trial ,” Lancet Oncology 25 ( 2024 ): 1202 – 1212 . 39146951 10.1016/S1470-2045(24)00338-3

23 D. Schadendorf , R. Dummer , A. Hauschild , et al., “ Health‐Related Quality of Life in the Randomised KEYNOTE‐002 Study of Pembrolizumab Versus Chemotherapy in Patients With Ipilimumab‐Refractory Melanoma ,” European Journal of Cancer 67 ( 2016 ): 46 – 54 . 27596353 10.1016/j.ejca.2016.07.018

24 D. J. Vaughn , J. Bellmunt , Y. Fradet , et al., “ Health‐Related Quality‐Of‐Life Analysis From KEYNOTE‐045: A Phase III Study of Pembrolizumab Versus Chemotherapy for Previously Treated Advanced Urothelial Cancer ,” Journal of Clinical Oncology 36 ( 2018 ): 1579 – 1587 . 29590008 10.1200/JCO.2017.76.9562

25 M. I. Milowsky , P. H. O'Donnell , C. J. Hoimes , et al., “ Patient‐Reported Outcomes in Patients With Advanced Urothelial Cancer Who Are Ineligible for Cisplatin and Treated With First‐Line Enfortumab Vedotin Alone or With Pembrolizumab ,” Journal of Clinical Oncology 42 ( 2024 ): 1403 – 1414 . 38215355 10.1200/JCO.23.01547 PMC11095879

26 T. Andre , M. Amonkar , J. M. Norquist , et al., “ Health‐Related Quality of Life in Patients With Microsatellite Instability‐High or Mismatch Repair Deficient Metastatic Colorectal Cancer Treated With First‐Line Pembrolizumab Versus Chemotherapy (KEYNOTE‐177): An Open‐Label, Randomised, Phase 3 Trial ,” Lancet Oncology 22 ( 2021 ): 665 – 677 . 33812497 10.1016/S1470-2045(21)00064-4

27 T. Leal , R. Kotecha , R. Ramlau , et al., “ Tumor Treating Fields Therapy With Standard Systemic Therapy Versus Standard Systemic Therapy Alone in Metastatic Non‐Small‐Cell Lung Cancer Following Progression on or After Platinum‐Based Therapy (LUNAR): A Randomised, Open‐Label, Pivotal Phase 3 Study ,” Lancet Oncology 24 ( 2023 ): 1002 – 1017 . 37657460 10.1016/S1470-2045(23)00344-3

28 R. Blanco , M. Dómine , J. L. González , et al., “ Pembrolizumab as First‐Line Treatment for Advanced NSCLC in Older Adults: A Phase II Clinical Trial Evaluating Geriatric and Quality‐Of‐Life Outcomes ,” Lung Cancer 183 ( 2023 ): 107318 . 37557022 10.1016/j.lungcan.2023.107318

29 T. Masuda , K. Fujitaka , T. Suzuki , et al., “ Phase 2 Study of First‐Line Pembrolizumab Monotherapy in Elderly Patients With Non‐Small‐Cell Lung Cancer Expressing High PD‐L1 ,” Thoracic Cancer 13 ( 2022 ): 1611 – 1618 . 35488720 10.1111/1759-7714.14428 PMC9161325

30 M. C. Garassino , S. Gadgeel , E. Esteban , et al., “ Patient‐Reported Outcomes Following Pembrolizumab or Placebo Plus Pemetrexed and Platinum in Patients With Previously Untreated, Metastatic, Non‐Squamous Non‐Small‐Cell Lung Cancer (KEYNOTE‐189): A Multicentre, Double‐Blind, Randomised, Placebo‐Controlled, Phase 3 Trial ,” Lancet Oncology 21 ( 2020 ): 387 – 397 . 32035514 10.1016/S1470-2045(19)30801-0

31 J. Mazieres , D. Kowalski , A. Luft , et al., “ Health‐Related Quality of Life With Carboplatin‐Paclitaxel or Nab‐Paclitaxel With or Without Pembrolizumab in Patients With Metastatic Squamous Non‐Small‐Cell Lung Cancer ,” Journal of Clinical Oncology 38 ( 2020 ): 271 – 280 . 31751163 10.1200/JCO.19.01348

32 F. Barlesi , E. B. Garon , D. W. Kim , et al., “ Health‐Related Quality of Life in KEYNOTE‐010: A Phase II/III Study of Pembrolizumab Versus Docetaxel in Patients With Previously Treated Advanced, Programmed Death Ligand 1‐Expressing NSCLC ,” Journal of Thoracic Oncology 14 ( 2019 ): 793 – 801 . 30711649 10.1016/j.jtho.2019.01.016

33 J. R. Brahmer , D. Rodríguez‐Abreu , A. G. Robinson , et al., “ Health‐Related Quality‐Of‐Life Results for Pembrolizumab Versus Chemotherapy in Advanced, PD‐L1‐Positive NSCLC (KEYNOTE‐024): A Multicentre, International, Randomised, Open‐Label Phase 3 Trial ,” Lancet Oncology 18 ( 2017 ): 1600 – 1609 . 29129441 10.1016/S1470-2045(17)30690-3

34 R. Dent , J. Cortés , L. Pusztai , et al., “ Neoadjuvant Pembrolizumab Plus Chemotherapy/Adjuvant Pembrolizumab for Early‐Stage Triple‐Negative Breast Cancer: Quality‐Of‐Life Results From the Randomized KEYNOTE‐522 Study ,” JNCI Journal of the National Cancer Institute 116 ( 2024 ): 1654 – 1663 . 38913881 10.1093/jnci/djae129 PMC11461162

35 D. W. Cescon , P. Schmid , H. S. Rugo , et al., “ Health‐Related Quality of Life With Pembrolizumab Plus Chemotherapy vs Placebo Plus Chemotherapy for Advanced Triple‐Negative Breast Cancer: KEYNOTE‐355 ,” JNCI Journal of the National Cancer Institute 116 ( 2024 ): 717 – 727 . 38070159 10.1093/jnci/djad240

36 P. Schmid , O. Lipatov , S. A. Im , et al., “ Impact of Pembrolizumab Versus Chemotherapy on Health‐Related Quality of Life in Patients With Metastatic Triple‐Negative Breast Cancer: Results From the Phase 3 Randomised KEYNOTE‐119 Study ,” European Journal of Cancer 195 ( 2023 ): 113393 . 37976633 10.1016/j.ejca.2023.113393

37 D. O'Malley , D. M. O'Malley , G. M. Bariani , et al., “ Health‐Related Quality of Life With Pembrolizumab Monotherapy in Patients With Previously Treated Advanced Microsatellite Instability High/Mismatch Repair Deficient Endometrial Cancer in the KEYNOTE‐158 Study ,” Gynecologic Oncology 166 ( 2022 ): 245 – 253 . 35835611 10.1016/j.ygyno.2022.06.005

38 N. Shaverdian , A. E. Lisberg , K. Bornazyan , et al., “ Previous Radiotherapy and the Clinical Activity and Toxicity of Pembrolizumab in the Treatment of Non‐Small‐Cell Lung Cancer: A Secondary Analysis of the KEYNOTE‐001 Phase 1 Trial ,” Lancet Oncology 18 ( 2017 ): 895 – 903 . 28551359 10.1016/S1470-2045(17)30380-7 PMC5538772

39 D. B. Johnson , S. Chandra , and J. A. Sosman , “ Immune Checkpoint Inhibitor Toxicity in 2018 ,” JAMA 320 ( 2018 ): 1702 – 1703 . 30286224 10.1001/jama.2018.13995

40 J. C. Pritchett , D. Patt , G. Thanarajasingam , A. Schuster , and C. Snyder , “ Patient‐Reported Outcomes, Digital Health, and the Quest to Improve Health Equity ,” American Society of Clinical Oncology Educational Book 43 ( 2023 ): e390678 . 37290027 10.1200/EDBK_390678

41 X. Yao , J. Zhou , L. Song , Y. Ren , P. Hu , and D. Liu , “ A Model‐Based Meta Analysis Study of Sodium Glucose Co‐Transporter‐2 Inhibitors ,” CPT: Pharmacometrics & Systems Pharmacology 12 ( 2023 ): 487 – 499 . 36890732 10.1002/psp4.12934 PMC10088079

42 M. Calvert , J. Blazeby , D. G. Altman , et al., “ Reporting of Patient‐Reported Outcomes in Randomized Trials: The CONSORT PRO Extension ,” JAMA 309 ( 2013 ): 814 – 822 . 23443445 10.1001/jama.2013.879

43 M. Calvert , D. Kyte , R. Mercieca‐Bebber , et al., “ Guidelines for Inclusion of Patient‐Reported Outcomes in Clinical Trial Protocols: The SPIRIT‐PRO Extension ,” JAMA 319 ( 2018 ): 483 – 494 . 29411037 10.1001/jama.2017.21903

44 Food and Drug Administration , “Patient‐Focused Drug Development: Collecting Comprehensive and Representative Input. Guidance for Industry. Food and Drug Administration Staff, and Other Stakeholders, 4–5,” ( 2020 ).

45 J. Z. Musoro , C. Coens , M. A. G. Sprangers , et al., “ Minimally Important Differences for Interpreting EORTC QLQ‐C30 Change Scores Over Time: A Synthesis Across 21 Clinical Trials Involving Nine Different Cancer Types ,” European Journal of Cancer 188 ( 2023 ): 171 – 182 . 37257278 10.1016/j.ejca.2023.04.027

46 M. Zettler , E. Basch , and C. Nabhan , “ Surrogate End Points and Patient‐Reported Outcomes for Novel Oncology Drugs Approved Between 2011 and 2017 ,” JAMA Oncology 5 ( 2019 ): 1358 – 1359 . 31268471 10.1001/jamaoncol.2019.1760 PMC6613294