Analysis of quality-adjusted survival time without symptoms or toxicity for pembrolizumab plus chemotherapy as treatment for previously untreated participants with advanced or metastatic esophageal cancer

Esophageal cancer is the sixth-leading cause of death from cancer worldwide, with a 5-year relative survival rate of approximately 25% [1, 2]. For several decades, the standard-of-care for untreated locally advanced or metastatic esophageal cancer has been chemotherapy combined with a fluoropyrimidine (fluorouracil or capecitabine) and a platinum agent (oxaliplatin or cisplatin), but overall survival (OS) remains poor [3]. Treatments that prolong OS for patients who have advanced esophageal cancer are necessary.

The KEYNOTE-590 (NCT03189719) trial was the first randomized phase 3 trial to show clinically meaningful and statistically significant improvement in OS (hazard ratio [HR], 0.73; 95% CI, 0.62–0.86; p < 0.0001) and progression-free survival (PFS; HR, 0.65, 95% CI, 0.55–0.76; p < 0.0001) with use of a programmed cell death protein 1 (PD-1) inhibitor combined with chemotherapy, compared with chemotherapy alone, in participants with previously untreated advanced esophageal cancer and Siewert type 1 gastroesophageal junction (GEJ) cancer [4]. A manageable safety profile was observed for pembrolizumab combined with chemotherapy compared with chemotherapy alone, and health-related quality of life (HRQOL) was maintained over time and was similar in the two treatment groups [4, 5].

Increasingly, regulatory bodies are demanding evaluation of the effect of a treatment on patient quality of life (QOL) [6]. The American Society of Clinical Oncology framework is focused on a net health benefit that measures the impact of an intervention by subtracting the division of incremental cost by the opportunity cost threshold from the incremental quality-adjusted life year gains. The Q-TWiST method is also used to assess treatment by integrating survival, toxicity, and QOL into a single measure, and it provides an analysis comparable to a net health benefit. With its relative simplicity and clinical relevance, Q-TWiST can better inform the shared decision making between clinicians and patients to assess the benefit of a treatment [7]. Analysis using the Q-TWiST method was initially adopted for adjuvant therapies in oncology, and it was widely used across different types of cancer over the past years [7‐11]. For instance, Q-TWiST was used to evaluate the benefits of pazopanib versus sunitinib in patients with renal cell carcinoma, demonstrating a slightly longer Q-TWiST, primarily due to a reduced length of time with grade 3 or 4 toxicity [11]. This type of analysis complements traditional outcomes generated from clinical trials by providing evidence used to evaluate the benefit/risk profile for patients. We performed a Q-TWiST analysis using data from the KEYNOTE-590 study to examine the benefit/risk profile of adding pembrolizumab to standard-of-care chemotherapy for participants who have advanced esophageal cancer. To our knowledge, this is the first Q-TWiST study conducted for pembrolizumab plus chemotherapy in advanced esophageal cancer across different histologies.

Using the Kaplan-Meier curve for OS, PFS, and TOX for each treatment group separately, the time spent in each of the three Q-TWiST health states is illustrated by the shaded area between the curves across four populations (Fig. 1).

At month 30, the mean duration was longer in the TOX state (1.57 months; 95% CI, 0.76–2.39) and the TWiST state (1.39 months; 95% CI, 0.04–2.84) and shorter in the REL state (0.30 month; 95% CI, − 2.07 to 1.38) for all randomly assigned participants in the pembrolizumab plus chemotherapy group compared with the chemotherapy alone group (Table 1).

We observed a similar trend in the participants with PD-L1 CPS ≥ 10 tumors and participants with ESCC. Notably, the mean duration of all three health states was longer in the participants with ESCC and a PD-L1 CPS ≥ 10 for the pembrolizumab plus chemotherapy group than in the chemotherapy alone group (Table 1).

Table 1 shows the restricted mean duration in the three different health states and restricted mean Q-TWiST at month 30 for the intention-to-treat population. After applying the pooled mean US health utility weights for each health state (Table 2), we saw that the Q-TWiST difference favored pembrolizumab plus chemotherapy compared with chemotherapy alone by 2.23 months for all randomly assigned participants (relative gain, 17.86%; Table 1; Fig. 2A).

We saw a similar trend in each subpopulation (PD-L1 CPS ≥ 10 tumors: 3.37 months [relative gain, 27.79%; Table 1; Fig. 2B]; ESCC: 2.29 months [relative gain, 18.12%; Table 1; Fig. 2C]; ESCC and a PD-L1 CPS ≥ 10: 3.87 months [relative gain, 33.47%; Table 1; Fig. 2D]).

In KEYNOTE-590, OS and PFS were better for participants treated with pembrolizumab plus chemotherapy compared with those who received only chemotherapy. However, use of pembrolizumab plus chemotherapy also led to increased toxicity. Thus, when evaluating treatment options, clinical efficacy, toxicity, and HRQoL should be considered.

Q-TWiST analyses are generally focused on clinical significance rather than on statistical significance, reporting effect sizes and the corresponding 95% CIs. In the current analysis, the mean difference and relative gain in quality-adjusted survival time between treatment groups were clinically meaningful and were substantial for all randomly assigned participants across the three subpopulations evaluated. The gain in quality-adjusted survival time was primarily due to a longer PFS duration. The numerically longer mean Q-TWiST observed for pembrolizumab plus chemotherapy versus for chemotherapy alone at month 30 supports the OS and PFS data observed in the KEYNOTE-590 trial [4], as well as observations from HRQoL analyses in which pembrolizumab plus chemotherapy did not negatively impact QoL compared with chemotherapy alone [5]. The 2.23 month gain in Q-TWiST indicates that patients may experience over two additional months of quality-adjusted and symptom-free survival compared to the control arm. In clinical practice, this translates to a meaningful extension of time during which patients are likely to remain symptom free and toxicity free, supporting continued functional independence and engagement in daily activities, potentially delaying transitions to more intensive care, and thereby helping clinicians and patients make more informed treatment decisions.

The relative Q-TWiST gain in the current analysis was 17.86% at a maximum follow-up of 30 months for all randomly assigned participants, a numerical improvement over the relative Q-TWiST gain (13%) reported for nivolumab plus chemotherapy versus chemotherapy alone at a maximum follow-up of 33 months for previously untreated participants with advanced gastric cancer, GEJ cancer, or esophageal adenocarcinoma [17]. In participants with ESCC, nivolumab plus ipilimumab and nivolumab plus chemotherapy both provided clinically meaningful improvement in Q-TWiST compared with chemotherapy alone (5.7 weeks and 7.6 weeks, respectively) after a minimum follow-up of 29 months [16]. The relative gain in quality-adjusted survival time of more than 15% is consistent with the relative gain of 17.0% to 20.0% reported for pembrolizumab monotherapy or in combination with chemotherapy compared with chemotherapy alone in participants with different tumor types [15, 18], and it is higher than the relative Q-TWiST gain reported across tumors and therapies in a previous systematic literature review [14]. In the systematic literature review that contextualizes and benchmarks Q-TWiST analysis [14], 51 articles with 81 distinct Q-TWiST comparisons were identified. As reported, the mean (median) relative Q-TWiST gains were 7.8% (7.2%) across all cancer types. Forty percent of the studies reported a relative Q-TWiST gain ≥ 10% (defined as clinically important difference) and 22.7% of the studies reported a relative Q-TWiST gain ≥ 15% (defined as clearly clinically important difference). The 10% and 15% thresholds were recommended in a previous systematic literature review conducted by Revicki, Feeny, Hunt, and Cole [7]. However, the authors stated that these clinically important difference estimates should be interpreted with caution, pending confirmation in future studies by direct patient assessment of the clinically relevant health states for Q-TWiST. Additionally, one should make comparisons to the literature with caution, considering the differences in participant baseline characteristics, follow-up duration, tumor type, and treatment regimen. The relative gain in quality-adjusted survival time was more pronounced in participants with PD-L1 CPS ≥ 10 tumors and participants who had ESCC and a PD-L1 CPS ≥ 10, likely a result of the extended PFS benefit observed. In addition, the relative gain in quality-adjusted survival time obtained in the base case exceeded established thresholds reported in the literature of at least 15.0%, which is considered “clearly clinically important,” although these established thresholds are not specific to participants with advanced esophageal cancer [7]. With a relative gain of greater than 15%, pembrolizumab plus chemotherapy conferred a clearly clinically meaningful improvement in quality-adjusted survival time compared with chemotherapy alone. Q-TWiST analysis provides an assessment of the benefit–risk associated with cancer treatment. These findings continue to support the use of pembrolizumab plus chemotherapy for this population.

As observed from the sensitivity analyses, the relative gain in quality-adjusted survival time increased as follow-up time increased, indicating a better PFS benefit. Using standardized utilities as mentioned in the methods section, a lower relative gain in quality-adjusted survival time was observed compared with using US utilities, implying that the impact of combined TOX and REL outweighed the impact of TWiST. The utility threshold analysis showed that Utility_TOX had a more substantial impact on relative Q-TWiST gain than Utility_REL, as expected given the larger difference in restricted time spent in the TOX state compared with the REL state. With increasing Utility_TOX, the relative Q-TWiST gain increased across all three subpopulations, whereas the relative Q-TWiST gain increased with increasing Utility_REL for all subpopulations except ESCC PD-L1 CPS ≥ 10.

One of the limitations of the primary analysis is the TOX state only includes grade 3–5 AEs and ignores the potential impact of grade 1–2 AEs on QOL. To account for this limitation, we conducted a sensitivity analysis that defined TOX as grade 1–2 AEs. Additionally, post-progression toxicities were not included in the analysis and may also influence results. Furthermore, missing EQ-5D-5L data and the short follow-up time might introduce a potential bias when interpreting data. However, compliance with scheduled EQ-5D-5L assessments was high (approximately 90%) and well balanced between treatment arms. The utility threshold analysis, conducted to assess how different utility thresholds affect the estimated treatment effect, consistently demonstrated a benefit of pembrolizumab plus chemotherapy over chemotherapy alone. Immunotherapy is often associated with long-term survival and delayed treatment effects; therefore, with short follow-up times, using the traditional Q-TWiST method could contribute to underestimation of the quality-adjusted survival length with immunotherapies. Performing Q-TWiST analyses with longer follow-up times is expected to further reduce uncertainty. Short-term data may provide an incomplete assessment of the treatment effect; long-term data are needed to ensure the validity and reliability of these results. However, the sensitivity analysis showed the relevant treatment effect increased over time. Therefore, the Q-TWiST difference and relative gain would likely increase with longer follow-up.

Results of the current Q-TWiST analysis showed that use of pembrolizumab plus chemotherapy caused clinically meaningful and substantial improvement in quality-adjusted survival compared with chemotherapy alone in participants with untreated locally advanced or metastatic esophageal cancer. The relative gain in quality-adjusted survival time exceeded the established thresholds in the literature, defined as either “clinically important” or “clearly clinically important.” The sensitivity analyses showed results were similar regardless of different follow-up periods, utility values, and definitions of the TOX state. The results of the current analysis add evidence from participants’ perspectives to strengthen the benefit/risk profile of pembrolizumab plus chemotherapy for the assessed population. Future studies with longer follow-up will be important to validate the current findings considering the delayed treatment effects often associated with immunotherapy.