Projecting long-term clinical outcomes with larotrectinib compared with immune checkpoint inhibitors in metastatic nonsmall cell lung cancer and differentiated thyroid cancer

Abstract

BACKGROUND:

Larotrectinib is approved for patients with advanced NTRK gene fusion-positive solid tumors. Prior studies demonstrated promising results with larotrectinib compared with other systemic therapy. However, comparisons to checkpoint inhibitors, such as nivolumab or pembrolizumab, have not been done.

OBJECTIVE:

To estimate and compare expected life-years (LYs) and quality-adjusted LYs (QALYs) for patients with nonsmall cell lung cancer (NSCLC) eligible for larotrectinib vs patients with unknown NTRK gene fusion status on nivolumab or pembrolizumab. We also assessed patients with metastatic differentiated thyroid cancer (DTC), as pembrolizumab may be considered in certain circumstances.

METHODS:

We developed partitioned survival models to project long-term comparative effectiveness of larotrectinib vs nivolumab or pembrolizumab. Larotrectinib survival data were derived from an updated July 2021 analysis of 21 adult patients (≥18 years of age) with metastatic NTRK gene fusion-positive NSCLC and 21 with DTC. Survival inputs for nivolumab and pembrolizumab were obtained from published articles. Progression-free and overall survival were estimated using survival distributions (Exponential, Weibull, Log-logistic, and Log-normal). Exponential fits were chosen based on goodness-of-fit and clinical plausibility.

RESULTS:

In NSCLC, larotrectinib resulted in gains of 5.87 and 5.91 LYs compared to nivolumab and pembrolizumab, respectively, which translated to gains of 3.53 and 3.56 QALYs. In DTC, larotrectinib resulted in a gain of 5.23 LYs and 4.24 QALYs compared to pembrolizumab.

CONCLUSIONS:

In metastatic NSCLC and DTC, larotrectinib may produce substantial life expectancy and QALY gains compared to immune checkpoint inhibitors. Additional data with longer follow-up will further inform this comparison.

Plain language summary

Our study aimed to inform an answer to the following question: when a patient with metastatic nonsmall cell lung cancer or differentiated thyroid cancer is indicated for either larotrectinib or an immune checkpoint inhibitor, which treatment provides better clinical and health outcomes? In our simulation model that used a lifetime time horizon, patients treated with larotrectinib projected to have substantial health gains compared to an immune checkpoint inhibitor.

Implications for managed care pharmacy

In this modeling study, patients treated with larotrectinib projected to have gains of at least 5.87 life-years (3.53 quality-adjusted life-years) and 5.23 life-years (4.24 quality-adjusted life-years) compared to an immune checkpoint inhibitor in metastatic nonsmall cell lung cancer and differentiated thyroid cancer, respectively. In appropriate clinical situations based on a patient’s NTRK gene fusion status, treatment with larotrectinib may result in substantial health gains compared to an immune checkpoint inhibitor.

Targeted therapies in oncology have changed the treatment landscape as patient outcomes have resulted in dramatic improvement.¹ These targeted medications necessitate patients have the appropriate genomic profiling, which can lead to challenges in the regulatory approval process. One such challenge is sample size, as patients who are available to receive the treatment are fewer than a nontargeted therapy.² As a result, increasingly more oncology treatments are approved through single arm trials (SATs) without a trial control group.³ In the past 20 years, nearly a third of new oncology indications were based on SATs, and nearly all of SAT approvals were for treating locally advanced or metastatic disease.³

Larotrectinib is a type of targeted therapy that was approved for patients with advanced NTRK gene fusion-positive solid tumors in SATs.⁴ Previous studies of larotrectinib have shown promising long-term improvements against standard-of-care systemic therapies across several tumor types, including in nonsmall cell lung cancer (NSCLC) and differentiated thyroid cancer (DTC).^5-9 These studies used a modeling approach to overcome some of the inherent challenges with assessing targeted therapies, such as small sample sizes, because the prevalence of patients with NTRK gene fusion status is low in patients with lung and thyroid cancers (3.30% and 1.41%, respectively).¹⁰ Although the standard-of-care comparison group used estimates from patients with unknown NTRK gene fusion status, these studies helped signal whether larotrectinib would provide additional clinical benefit in specific tumor types for eligible patients.

Immune checkpoint inhibitors, including nivolumab and pembrolizumab, have also demonstrated durable response in a subset of patients with previously treated metastatic cancers.^11,12 However, comparisons of larotrectinib to nivolumab and pembrolizumab have not been done. This study aimed to estimate and compare the life-years (LYs) and quality-adjusted LYs (QALYs) for (1) patients with NSCLC eligible for larotrectinib vs patients with unknown NTRK gene fusion status on nivolumab or pembrolizumab and (2) patients with metastatic DTC eligible for larotrectinib vs patients with unknown NTRK gene fusion status on pembrolizumab.

Methods

MODELING APPROACH

A partitioned survival modeling approach was used to estimate the potential long-term comparative effectiveness of larotrectinib vs nivolumab or pembrolizumab in patients with metastatic NSCLC or DTC (pembrolizumab only) who progressed after first-line treatment (Supplementary Figure 1 ^{(262.8KB, pdf)} , available in online article). The model comprised the following 3 health states: progression free, progressed, and death. Partitioned survival methods (ie, area under the curve) were used to estimate the proportion of patients in each health state at a given time point from progression-free survival (PFS) and overall survival (OS) curves. A lifetime time horizon was used to estimate long-term projections, which were discounted at a 3% rate. One-way sensitivity analysis was run on all available PFS, OS, and utility parameters based on 95% CIs if available and otherwise by ±10% to evaluate influential parameters to the model. Probabilistic sensitivity analysis was run where all aforementioned parameters were varied simultaneously with 10,000 simulations to create 95% credible intervals (2.5th and 97.5th percentiles) around the outcomes. Applicable portions of reporting the study followed The Consolidated Health Economic Evaluation Reporting Standards statement.¹³

LAROTRECTINIB SURVIVAL DATA SOURCES

Across the NSCLC and DTC models, we used July 2021 larotrectinib survival data that were assessed by an independent review committee for adult patients (aged ≥18 years) from the larotrectinib clinical trials program (NCT02122913, NCT02637687, and NCT02576431).¹⁴ The NSCLC model survival curves were derived from 21 patients with an overall response rate (ORR) of 85.7%, including 2 patients (9.5%) with complete response and 16 patients (76.2%) with partial response as best response. The DTC model was informed by 21 patients with metastatic DTC who had received available kinase inhibitor therapy and of whom 20 had previously received radioactive iodine. The ORR was 81.0%, of which 2 patients (9.5%) had complete response and 15 (71.4%) had partial response as best response.

NIVOLUMAB AND PEMBROLIZUMAB DATA SOURCES

Nivolumab and pembrolizumab were used in patients with metastatic NSCLC with unknown NTRK gene fusion status. Baseline characteristics across the trials for nivolumab, pembrolizumab, and larotrectinib were collected for comparison (Supplementary Table 1 ^{(262.8KB, pdf)} ). Trial outcomes that were used to inform the model from the clinical trials are in Supplementary Table 2 ^{(262.8KB, pdf)} . In the NSCLC model, PFS and OS Kaplan-Meier curves for both checkpoint inhibitors were taken from and digitized from their published, pivotal phase 3 multicenter trials that included the United States and led to drug approval by the US Food and Drug Administration.^11,12 The ORR for nivolumab was 19.2%, with 1.4% of patients having complete response and 17.8% having partial response. The ORR for pembrolizumab was 18.0%, with all responses categorized as partial response.

In DTC, there has been interest in assessing checkpoint inhibitors because of the immune system and inflammatory pathways in the progression of DTC.¹⁵ Of note, pembrolizumab was recently assessed in DTC in the phase 2 KEYNOTE-158 study, which enrolled patients (n = 103) with previously treated advanced disease.¹⁶ The ORR was 6.8% (n = 7) with 2 patients with complete response and 5 with partial response.

PARAMETRIC SURVIVAL CURVE FITS

The PFS and OS curves from updated data for larotrectinib and published trials for nivolumab and pembrolizumab were extrapolated using parametric curve fits (Exponential, Weibull, Log-logistic, Log-normal).¹⁷ After assessing visual fit of the extrapolated curves, clinical plausibility, and minimal Bayesian information criteria, exponential fits were applied to PFS and OS curves for larotrectinib, nivolumab, and pembrolizumab across both cancer models. US life tables were applied in order to prevent the possibility of higher survival age-based mortality rates in any treatment arm.¹⁸

HEALTH STATE UTILITY VALUES FOR ESTIMATION OF QALYs

Published literature were used to estimate health state utility weights for progression-free and postprogression health states (Table 1). For NSCLC, utilities were elicited from 100 members of the United Kingdom general public using the standard gamble approach.¹⁹ The utility was 0.65 (95% CI = 0.61-0.70) for progression-free and 0.47 (95% CI = 0.41-0.51) for progressive disease. For DTC, utilities were elicited from a time trade-off study by 100 members of the United Kingdom general public for radioactive iodine-refractory DTC.²⁰ The utility was 0.80 (95% CI = 0.77-0.84) for progression-free and 0.50 (95% CI = 0.45-0.56) for progressive disease. Across both models, on-treatment utility weights were estimated as the weighted average of the utility for those in progression-free, with and without responsive disease, based on the response rate for each treatment.

TABLE 1.

Utility Values Used in Models

	Progression-free/stable disease	Progressive disease
Nonsmall cell lung cancer, base case (95% CI)	0.65 (0.61-0.70)	0.47 (0.41-0.51)
Differentiated thyroid cancer, base case (95% CI)	0.80 (0.77-0.84)	0.50 (0.45-0.56)

Results

In the NSCLC model, larotrectinib use led to 4.36 LYs (95% CrI = 2.06-7.47) and 2.92 QALYs (95% CrI = 1.36-5.01) in the progression-free state for the patient’s lifetime (Table 2). This resulted in a lifetime incremental gain of 3.82 LYs (2.57 QALYs) and 3.78 LYs (2.54 QALYs) compared to pembrolizumab and nivolumab, respectively. Additionally, median PFS was 41.08 months, 4.62 months, and 5.08 months for larotrectinib, pembrolizumab, and nivolumab, respectively. Given the similar outcomes for pembrolizumab and nivolumab, the extrapolated PFS and OS curves appeared to overlap (Figure 1). In total (progression free and postprogression), patients on larotrectinib resulted in 7.33 LYs (95% CrI = 3.28-11.42) and 4.33 QALYs (95% CrI = 2.16-6.49). Compared to pembrolizumab and nivolumab, this led to an incremental gain of 5.91 LYs (3.56 QALYs) and 5.87 LYs (3.53 QALYs), respectively. Median OS was 79.39 months, 12.23 months, and 12.69 months for larotrectinib, pembrolizumab, and nivolumab, respectively.

TABLE 2.

Survival and Quality-Adjusted Survival Outcomes in Nonsmall Cell Lung Cancer

	Larotrectinib	Pembrolizumab	Nivolumab
Progression-free LYs (95% CrI)	4.36 (2.06-7.47)	0.54 (0.48-0.60)	0.58 (0.51-0.66)
Postprogression LYs (95% CrI)	2.97 (0.00-7.76)	0.89 (0.69-1.11)	0.88 (0.68-1.11)
Total LYs (95% CrI)	7.33 (3.28-11.42)	1.42 (1.24-1.64)	1.46 (1.28-1.68)
Progression-free QALYs (95% CrI)	2.92 (1.36-5.01)	0.35 (0.30-0.41)	0.38 (0.32-0.45)
Postprogression QALYs (95% CrI)	1.41 (0.00-3.71)	0.42 (0.32-0.53)	0.42 (0.32-0.54)
Total QALYs (95% CrI)	4.33 (2.16-6.49)	0.77 (0.67-0.89)	0.80 (0.70-0.92)

CrI = credible interval; LY = life-year; QALY = quality-adjusted life-year.

FIGURE 1.

Extrapolated PFS and OS in Nonsmall Cell Lung Cancer

In the DTC model, the larotrectinib arm resulted in 5.38 LYs (95% CrI = 2.79-8.68) and 4.56 QALYs (95% CrI = 2.35-7.40) in the progression-free state for the patient’s lifetime (Table 3). This resulted in a lifetime gain of 4.54 LYs (3.89 QALYs) compared to pembrolizumab. Median PFS was 52.4 months and 7.2 months for larotrectinib and pembrolizumab, respectively (Figure 2). For OS, treatment with larotrectinib resulted in 9.30 LYs (95% CrI = 4.77-12.25) and 6.53 QALYs (95% CrI = 3.75-8.97). Compared to pembrolizumab, this led to an additional 5.23 LYs (4.24 QALYs). Median OS was 78.95 months and 38.1 months for larotrectinib and pembrolizumab, respectively.

TABLE 3.

Survival and Quality-Adjusted Survival Outcomes in Differentiated Thyroid Cancer

	Larotrectinib	Pembrolizumab
Progression-free LYs (95% CrI)	5.38 (2.79-8.68)	0.84 (0.69-1.02)
Postprogression LYs (95% CrI)	3.93 (0.00-8.08)	3.24 (2.38-4.24)
Total LYs (95% CrI)	9.30 (4.77-12.25)	4.07 (3.24-5.06)
Progression-free QALYs (95% CrI)	4.56 (2.35-7.40)	0.67 (0.55-0.83)
Postprogression QALYs (95% CrI)	1.96 (0.00-4.70)	1.62 (0.80-2.64)
Total QALYs (95% CrI)	6.53 (3.75-8.97)	2.29 (1.46-3.31)

CrI = credible interval; LY = life-year; QALY = quality-adjusted life-year.

FIGURE 2.

Extrapolated PFS and OS in Differentiated Thyroid Cancer

In the one-way sensitivity analysis across both cancer models, the most influential parameter on QALYs gained tended to be the PFS and OS input estimates for larotrectinib, followed by the utility estimates for the progressive and progression-free health states (Supplementary Figures 2-4 ^{(262.8KB, pdf)} ). The response rates and comparator PFS and OS input estimates were less influential across all models.

Discussion

In this analysis, larotrectinib clinical trial data were used to estimate potential long-term outcomes compared to nivolumab and pembrolizumab in NSCLC and pembrolizumab in DTC. Using partitioned survival methods allowed for extrapolation of available data for all therapies and comparisons within the same analysis. As eligible patients who have failed prior therapy in metastatic NSCLC may have to consider larotrectinib or a checkpoint inhibitor, and because a comparison has not been studied elsewhere, this current study provides evidence that larotrectinib may provide potentially substantial gains in terms of LYs and QALYs. Furthermore, as pembrolizumab may be used in the near future for advanced DTC, this study comparing larotrectinib to pembrolizumab resulted in potentially favorable results for larotrectinib in eligible patients.

Although it is difficult to validate the larotrectinib results of this study with real world–evidence studies, as individual tumor types have not been assessed with larotrectinib use, it is possible with pembrolizumab and nivolumab, as observational studies have been conducted with these therapies in patients with metastatic NSCLC who have failed prior therapies. In studies that used electronic health records, median PFS for second-line nivolumab ranged from 2.3 to 3.5 months and median OS ranged from 7.4 to 12.2 months.^21,22 In patients who used pembrolizumab as second line in metastatic NSCLC, median PFS ranged from 3.0 to 3.7 months and median OS ranged from 9.3 to 12.0 months.²¹ These estimates were in line with our modeled results of approximately 5 months and 12 months for PFS and OS in both pembrolizumab and nivolumab, respectively. Similar real world–evidence studies would help corroborate the results for larotrectinib in this study, but in the absence of such studies, the comparable results for the checkpoint inhibitors indicate the modeling methods used here may yield meaningful clinical outcomes for larotrectinib.

LIMITATIONS

There were limitations to the current study. In the larotrectinib arm, there was a small sample size that informed the parametric curves to the model. However, sensitivity analyses revealed that estimates for larotrectinib PFS were robust. As data mature, the current analysis should be replicated to provide more precise estimates and to lessen the limitation with the currently available data. Additionally, as this study used a naive direct approach, adjustment or matching for baseline differences across trial populations were not possible and thus could impact the results. Of note, patients treated with larotrectinib had NTRK gene fusions, whereas the status for patients treated with pembrolizumab was unknown. Although we conducted one-way and probabilistic sensitivity analyses to assess our results, some uncertainty remained given that the clinical inputs for the model were based on clinical trials and not real-world data. Finally, our study did not consider safety outcomes.

As targeted therapies in oncology become more common, issues, such as small sample size and single arm treatments, in clinical trials will continue to be a challenge. In such cases, methods, such as naive direct comparisons, may be the only viable approaches to help guide clinical decision-making in the absence of direct comparative evidence. However, results using these methods need to be interpreted carefully, and further validation is needed with supplemental studies, such as using electronic health record or registry data to allow for adjusting of baseline characteristics.

Conclusions

In this modeling study, patients on larotrectinib were projected to have gains in LYs and QALYs compared to pembrolizumab and nivolumab in metastatic NSCLC and pembrolizumab in metastatic DTC. These projections need to be taken carefully as they were the result of naive direct comparisons in patients with NTRK gene fusion-positive for larotrectinib \and unknown NTRK gene fusion for pembrolizumab and nivolumab. Although future studies using more mature clinical trial data or real-world evidence will help confirm the results of this study, our modeling results showed promising long-term clinical outcomes in patients treated with larotrectinib.

DATA AVAILABILITY

We developed a decision analytic model. There are no patient data.

Funding Statement

This study was funded by Bayer U.S. LLC. The funder reviewed and approved the manuscript for publication, but otherwise had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation or decision to submit the manuscript for publication.