Pembrolizumab plus chemotherapy with or without bevacizumab for persistent, recurrent or metastatic cervical cancer

Cervical cancer is defined as persistent when it does not respond to treatment, recurrent when it has returned after treatment and metastatic when it has spread beyond the cervix to other places in the body. The committee noted that persistent, recurrent or metastatic cervical cancer has a poor prognosis and leads to a substantial disruption to quality of life. There are limited effective treatment options available for this type of cancer. The main aim of treatment is to relieve symptoms and improve quality of life, and to extend life if possible. The committee concluded that there is a high disease burden for people with persistent, recurrent or metastatic cervical cancer. It also concluded that a treatment which can prolong life but also improve quality of survival by managing symptoms would be welcome.

The clinical evidence was based on KEYNOTE‑826, a phase 3, double-blind randomised placebo-controlled trial in people with recurrent, persistent or metastatic cervical cancer. KEYNOTE‑826 compared pembrolizumab plus chemotherapy with or without bevacizumab against placebo plus chemotherapy with or without bevacizumab as a first-line treatment. In line with the marketing authorisation, the company presented efficacy data for people whose tumours express PD‑L1 with a combined positive score (CPS) of at least 1. In the original NICE technology appraisal guidance on pembrolizumab plus chemotherapy with or without bevacizumab for persistent, recurrent or metastatic cervical cancer (from now TA885), the committee considered the interim analysis from the trial. For this rapid review, the company provided data from the final analysis of KEYNOTE‑826. This included 17 months of additional follow-up data, with median overall survival in the pembrolizumab group now reached. There was a clinically meaningful and statistically significant benefit for the pembrolizumab group compared with the placebo group for both progression-free survival (PFS) and overall survival (OS). Median PFS was 10.5 months in the pembrolizumab group and 8.2 months in the placebo group (hazard ratio [HR] 0.58, 95% confidence interval [CI] 0.47 to 0.71). This was a slight improvement on the hazard ratio from the interim analysis (HR 0.62, 95% CI 0.50 to 0.77). Median OS was 28.6 months in the pembrolizumab group and 16.5 months in the placebo group. (HR 0.60, 95% CI 0.49 to 0.74). This was also a slight improvement on the interim analysis (HR 0.64, 95% CI 0.50 to 0.81). The EAG commented that all reported endpoints from the final analyses were similar to the results from the interim analyses. It noted that the treatment effect sizes were generally similar in size, with greater precision because of the additional observed events. The committee concluded that the more mature data from KEYNOTE‑826 was consistent with the interim analysis and strengthened the clinical evidence for pembrolizumab.

The EAG questioned whether the trial data was generalisable to the NHS. It highlighted that clinical expert advice presented by the company estimated that people who have standard care could expect to live for 3 to 15 months. But median OS in the placebo arm in KEYNOTE‑826 was 16.5 months. Also, some people in the trial had subsequent immuno-oncology treatments that are not available on the NHS. The committee accepted that the subsequent treatments may have extended survival. It also considered that the difference in survival may have been because people in clinical trials tend to be fitter than people in clinical practice. The company said that clinical expert opinion had confirmed that the trial results were generalisable to the NHS population. The Cancer Drugs Fund lead said that real world outcomes often do not match those seen in clinical trials. There are several reasons for this (for example, people with additional health complications are typically excluded from trials). But he said there were no particular concerns about generalisability for this appraisal. He also noted that having a fitter trial population would lead to better outcomes in both arms of the trial, not just the pembrolizumab arm. The committee was reassured that the OS outcomes in the placebo group of KEYNOTE‑826 were consistent with those previously seen in GOG 240. This was a randomised phase 3 trial in people with persistent, recurrent or metastatic cervical cancer. The committee concluded that the results from KEYNOTE‑826 were broadly generalisable to the NHS and appropriate for decision making.

The company presented a 3‑state Markov state transition model to estimate the cost effectiveness of pembrolizumab plus chemotherapy compared with chemotherapy (both with or without bevacizumab). The 3 health states were PFS, progressed disease and death. In TA885, the company explained that there was not enough OS data for people whose cancer responded to treatment. So, the data on OS was not mature enough to accurately model long-term survival, particularly in people whose cancer responded completely. Taking these factors into account, the committee for TA885 agreed that a state transition model (STM) was more appropriate than a partitioned survival model (PSM). This was because an STM uses a structural link between PFS and OS rather than directly extrapolating OS data. But the committee noted that the appropriate model structure may change when there was more data. In its submission for this rapid review, the company said that the OS data still did not fully reflect the survival benefit for people whose cancer responded well to treatment. So, its base case still used the STM. At the EAG's request, the company also provided a PSM. But the EAG acknowledged that the results generated by the PSM were not clinically plausible. The EAG said that the most logical extrapolations resulted in an unrealistic crossing of OS and PFS curves. This was because OS data was too immature to capture the plateau trend seen in the PFS curve. But the EAG noted that the cost-effectiveness results produced using a PSM remained relatively robust and were unlikely to be a key driver of uncertainty. The committee agreed that, given the implausible results generated by the PSM, the STM was still the most appropriate modelling approach. But it acknowledged the remaining uncertainty about the long-term benefits of pembrolizumab, particularly among people whose cancer responds well to treatment.

In TA885, the committee noted that the long-term projections for PFS were highly uncertain. Given the structural link between PFS and OS in the model structure (see section 3.4), this also drove uncertainty in the long-term projections for OS. It noted that these uncertainties may be reduced when there was longer follow-up data from KEYNOTE‑826. The final analysis has provided this data, and the company said that the range of appropriate extrapolation curves for PFS has been narrowed, so reducing this uncertainty. The company's base case selected a 3‑knot hazard model, which gave similar estimates to those previously used in TA885. The company also examined 1‑knot and 2‑knot models in sensitivity analyses. The EAG was 'generally satisfied' with the company's extrapolation approach, agreeing there was a good visual and statistical fit to the updated data. But it noted that much of the modelled incremental benefit associated with pembrolizumab was in the extrapolated portion of the survival curves. So, it thought that some uncertainty remained about the long-term survival benefit with pembrolizumab, particularly among people whose cancer responds well to treatment. The committee concluded that the company's extrapolation of the survival curves was appropriate, noting that some uncertainty remained.

NICE's guide to the methods of technology appraisal notes that above a most plausible incremental cost-effectiveness ratio (ICER) of £20,000 per quality-adjusted life year (QALY) gained, decisions about the acceptability of a technology as an effective use of NHS resources will take into account the degree of certainty around the ICER. The committee will be more cautious about recommending a technology if it is less certain about the ICERs presented. The committee accepted that the amount of uncertainty had been reduced since TA885 because of the additional 17 months of follow-up data. But it noted that there was still uncertainty about the long-term survival benefit of pembrolizumab, particularly among people whose cancer responds well to treatment (see section 3.5). The committee considered the remaining uncertainty and the greater weight given to QALYs at the end of life (see section 3.6). It concluded that the maximum acceptable ICER would be substantially below £50,000 per QALY gained.

The committee's preferred assumptions from TA885 were implemented into the company's base case for this rapid review. They included a treatment effect waning from 3 years to 5 years after stopping pembrolizumab with a 2‑year stopping rule. The company's base-case ICER for pembrolizumab plus chemotherapy compared with placebo plus chemotherapy (both with or without bevacizumab), and ICERs from scenario analyses exploring key assumptions, were all substantially below the acceptable level of £50,000 per QALY gained (see section 3.7). Because of confidential commercial arrangements for pembrolizumab, bevacizumab and postprogression therapies, the ICERs are confidential and cannot be reported here. The EAG was satisfied with the company's analyses and did not provide any additional scenarios. Because the ICERs were below the agreed level, the committee concluded that pembrolizumab plus chemotherapy with or without bevacizumab could be recommended for routine use.

The committee concluded that the more mature data from KEYNOTE‑826 had strengthened the clinical evidence for pembrolizumab plus chemotherapy with or without bevacizumab. The most plausible ICERs were within the range usually considered a cost-effective use of resources when the end of life modifier was applied. So, pembrolizumab plus chemotherapy with or without bevacizumab is recommended as an option for treating persistent, recurrent or metastatic cervical cancer in adults whose tumours express PD‑L1 with a CPS of at least 1.