Erlotinib monotherapy for maintenance treatment of non-small-cell lung cancer

The manufacturer approached the decision problem by providing clinical and cost-effectiveness evidence for erlotinib maintenance monotherapy compared with best supportive care in patients with stage 3B or stage 4 squamous or non-squamous non-small-cell lung cancer who had stable disease after treatment with standard platinum-based first-line chemotherapy. Best supportive care included palliative radiotherapy, corticosteroids, analgesia and other symptomatic treatments and watchful waiting alone. In the economic evaluation the manufacturer provided combined and separate analyses for patients with squamous and non-squamous disease. The group of patients with non-squamous disease was further divided into 2 analyses: firstly those who were not eligible for pemetrexed maintenance therapy (that is, patients who received pemetrexed in combination with cisplatin as first-line treatment), with best supportive care as the comparator; and secondly those who were eligible for pemetrexed maintenance therapy (that is, patients who have received first-line treatment with platinum-based chemotherapy which did not include pemetrexed), with pemetrexed as the comparator. The manufacturer noted a lack of head-to-head clinical evidence comparing erlotinib with pemetrexed.

At the time of the manufacturer's original evidence submission, the marketing authorisation for erlotinib had not been granted and the population who would be covered by the marketing authorisation was unclear. Subsequently, erlotinib received a marketing authorisation for the maintenance treatment of a subgroup of patients with locally advanced or metastatic non-small-cell lung cancer, that is, patients with stable disease after standard platinum-based first-line chemotherapy. Because of this, some of the evidence included in the manufacturer's original submission was not relevant and further evidence for the population covered by the marketing authorisation was required. The manufacturer provided additional clinical and cost-effectiveness evidence for patients with stable disease after first-line chemotherapy in response to consultation on the first appraisal consultation document.

The key evidence submitted by the manufacturer for the clinical effectiveness of erlotinib came from a randomised double-blind controlled trial comparing erlotinib with placebo in patients with advanced or metastatic non-small-cell lung cancer whose disease had not progressed following platinum-based first-line chemotherapy (the SATURN trial). This population included patients whose disease was either stable after first-line chemotherapy, or had responded to first-line chemotherapy. Stable disease was defined according to the Response Evaluation Criteria in Solid Tumours (RECIST) as tumour shrinkage that is not sufficient to be classed as a partial response and tumour increase that is not sufficient to be classed as progressive disease.

Patients were included in the SATURN trial if their disease had not progressed after 4 cycles of a standard, platinum-based chemotherapy doublet (2 chemotherapy drugs, 1 of which is platinum based), if they had an Eastern Cooperative Oncology Group (ECOG) performance status of 0 to 1 and their life expectancy was at least 12 weeks. The primary outcome of the trial was progression-free survival, defined as the time between randomisation and the date of the first documented disease progression, or death from any cause. Secondary outcomes included overall survival (defined as the time between randomisation and death), time to disease progression, response rates (assessed by the RECIST criteria) and quality of life (assessed by the Functional Assessment of Cancer Therapy – Lung [FACT-L] questionnaire). RECIST criteria were assessed by computed tomography (CT) scans every 6 weeks.

The SATURN trial included 889 patients, of whom 487 (55%) had stable disease after first-line chemotherapy. Of the patients with stable disease, approximately 30% had an ECOG performance status of 0 and 70% had an ECOG performance status of 1 at study entry, 20% had never smoked and 61% had non-squamous disease. Fifty percent of patients were tested for EGFR mutation status and of these 11% had activated EGFR mutations. The most common first-line treatments for patients in the whole trial population were gemcitabine plus carboplatin (28%), gemcitabine plus cisplatin (26%), and paclitaxel plus carboplatin (19%). None of the patients with non-squamous disease received combination chemotherapy with cisplatin and pemetrexed, now the most commonly prescribed first-line regimen in UK clinical practice for this histological subtype of non-small-cell lung cancer. Most patients (48%) were from Eastern Europe, 21% were from south-east Asia and 1% were from the UK. The mean age of patients was 60 years. The proportion of patients who had at least 1 post-study treatment was 72% in the placebo group and 71% in the erlotinib group, with 21% of patients in the placebo group and 11% in the erlotinib group receiving subsequent treatment with a tyrosine kinase inhibitor (such as erlotinib or gefitinib). The proportion of patients with stable disease who had further systemic therapy after the study was 63% in the placebo group and 61% in the erlotinib group, with 21% of patients in the placebo group and 9% in the erlotinib group receiving subsequent treatment with either erlotinib or gefitinib. Of the 50% of patients with stable disease who were tested, the incidence of activated EGFR mutation was 6% in both the placebo and the erlotinib groups. The manufacturer stated that there were no imbalances between treatment arms when clinical, molecular, or geographical parameters were considered or when prior radiotherapy, subsequent systemic treatments and time to start of investigational treatment were analysed.

The manufacturer did not provide separate demographic analyses for the squamous and non-squamous disease groups in response to consultation on the first appraisal consultation document. However, during consultation on the second appraisal consultation document, the manufacturer indicated that although the baseline characteristics of known prognostic significance in the stable, squamous disease group were reasonably balanced between the erlotinib and best supportive care arms, this was not the case for the non-squamous disease group. Data from the SATURN trial showed that for the population with stable non-squamous disease, 30% of patients in the erlotinib group and 38% in the placebo group had an ECOG status of 0, and 25% of patients in the erlotinib group and 31% in the placebo group had never smoked. The manufacturer stated that because of these imbalances, the true overall survival benefit from erlotinib in the non-squamous disease group is likely to be confounded in favour of the best supportive care group.

For patients with stable disease after first-line chemotherapy, median progression-free survival was 12.1 weeks in the erlotinib group compared with 11.3 weeks in the placebo group (hazard ratio [HR] 0.68; 95% confidence interval [CI] 0.56 to 0.83, p<0.0001). Median overall survival was 11.9 months in the erlotinib group compared with 9.6 months in the placebo group, a difference of 2.3 months (HR 0.72; 95% CI 0.59 to 0.89, p=0.0019). The increase in progression-free survival associated with erlotinib was similar in patients with squamous disease (HR 0.69; 95% CI 0.51 to 0.93, n=190) and non-squamous disease (HR 0.69; 95% CI 0.54 to 0.87, n=297). The gain in median overall survival associated with erlotinib for patients with squamous disease was 3.0 months (HR 0.67; 95% CI 0.48 to 0.92) and 3.1 months for patients with non-squamous disease (HR 0.76; 95% CI 0.59 to 1.00). There were no statistically significant differences between the erlotinib and placebo groups in quality-of-life measures. In response to consultation on the first appraisal consultation document, the manufacturer provided estimates for the mean overall survival benefit associated with erlotinib for patients with squamous and non-squamous disease (4.5 months and 4.2 months respectively). The mean overall survival estimate for the whole stable disease population was 3.3 months.

Subgroup analyses were carried out by the manufacturer in patients with stable disease after first-line chemotherapy by EGFR immunohistochemistry (IHC) status, stage of disease, first-line chemotherapy regimen received, ECOG performance status, smoking status, geographical region, age, race and gender. Erlotinib was associated with a greater median overall survival benefit for patients with EGFR IHC-positive tumours (HR 0.65; 95% CI 0.51 to 0.88, p value not reported), for patients who had never smoked (HR 0.56; 95% CI 0.32 to 0.97, p value not reported), and for women (HR 0.55; 95% CI 0.35 to 0.87, p value not reported) compared with the whole stable disease population. In covariate analyses from the manufacturer of the whole stable SATURN population, south-east Asian patients had a greater progression-free survival benefit from erlotinib than white patients (covariate effect HR 0.78; p=0.0214), as did patients who had never smoked compared with current smokers (covariate effect HR 1.46; p=0.0003). In addition, the small subgroup of patients with activated EGFR mutations were also shown to gain substantially more benefit from erlotinib than the trial population as a whole (HR for progression-free survival 0.23; 95% CI 0.12 to 0.45).

During consultation on the second appraisal consultation document, the manufacturer provided adjusted analyses comparing erlotinib with best supportive care in patients with stable, non-squamous disease from the SATURN trial. The analysis suggested that when the data were adjusted for good ECOG performance status and smoking status (never smoked), patients with these characteristics had an improved rate of overall survival after treatment with erlotinib compared with the whole stable, non-squamous disease population (HR 0.71; 95% CI 0.54 to 0.93). The estimate of overall survival was even higher (HR 0.63; 95% CI 0.41 to 0.96) when the analysis was repeated only in patients without an EGFR mutation (that is, the patients who are likely to receive erlotinib in UK clinical practice due to the growing use of gefitinib in patients with an EGFR mutation). In light of these analyses, the manufacturer stated that the overall survival benefit from erlotinib in the non-squamous disease population may have been underestimated in the SATURN trial.

In response to the second appraisal consultation document, the manufacturer provided patient characteristics for the squamous disease group to allay the Committee's concerns that the population in the SATURN trial may not be representative of patients seen in the UK. In particular, they showed that only 0.5% of patients with squamous disease had an EGFR mutation, 6.9% had never smoked and 7.9% were Asian. Given the relatively low incidence of patients with these prognostic factors, the manufacturer stated that they would not have a large impact on the overall survival benefit observed in the SATURN trial for the squamous disease group relative to what would be achieved in UK clinical practice.

The most common adverse events associated with erlotinib for the whole SATURN population were rash (49% in the erlotinib group and 6% in the placebo group) and diarrhoea (20% in the erlotinib group and 5% in the placebo group). In the stable disease subgroup, more patients in the erlotinib group had an adverse event of any kind than in the placebo group (78% compared with 58%). There were 23 deaths in the erlotinib group and 22 in the placebo group during the active treatment phase.

In the original submission, the manufacturer carried out an indirect analysis of erlotinib and pemetrexed in patients with non-squamous disease using data from a randomised controlled trial of pemetrexed maintenance treatment versus placebo in patients with locally advanced or metastatic non-small-cell lung cancer (the JMEN trial). This analysis included patients with stable disease but also included patients whose disease had responded to first-line treatment with platinum-based chemotherapy. The latter group are not covered by the marketing authorisation for erlotinib. In the additional evidence submission, received during consultation on the first appraisal consultation document, the manufacturer stated that an indirect comparison of the SATURN and JMEN trials was not possible because no data on the efficacy of pemetrexed in patients with stable, non-squamous disease from the JMEN trial were publicly available. The manufacturer also stated that an indirect analysis was not considered appropriate because of differences in the populations in the 2 trials.

In the original submission, the manufacturer submitted economic analyses for 3 different patient populations, 2 of which included patients who were not covered by the marketing authorisation, that is, patients whose disease had responded to treatment. The only economic analysis that reflected the population covered by the marketing authorisation was the comparison of erlotinib with best supportive care in patients with stable disease. In response to consultation on the first appraisal consultation document, the manufacturer submitted 4 new economic analyses that included the population covered by the marketing authorisation:

• erlotinib compared with best supportive care in all patients with stable disease

• erlotinib compared with best supportive care in patients with stable, squamous disease

• erlotinib compared with best supportive care in patients with stable, non-squamous disease who were not eligible for pemetrexed maintenance treatment

• erlotinib compared with pemetrexed in patients with stable, non-squamous disease who were eligible for pemetrexed maintenance treatment.

The manufacturer's new economic analyses used a model with a cycle length of 1 month and a 15-year time horizon. The model included 3 health states: progression-free survival, progressed (defined as the time from first treatment relapse until death), and death. All patients were assumed to start in the progression-free survival health state (after first-line chemotherapy). At the end of each cycle, they could remain in this state, move to the progressed health state or die.

For the comparison of erlotinib with best supportive care, the risks of disease progression were taken from the SATURN trial. For the comparison of erlotinib with pemetrexed, the base case assumed equal efficacy of pemetrexed and erlotinib. Additional relative efficacy scenarios were modelled, ranging from assuming greater efficacy of pemetrexed (in progression-free survival and overall survival) to assuming greater efficacy of erlotinib.

In the new economic analyses, the manufacturer used the same utility values as those used in NICE's technology appraisal guidance on pemetrexed for the maintenance treatment of non-small-cell lung cancer. The utility values for the progression-free survival health state were 0.6732 for the erlotinib group and 0.6628 for the placebo group. The utility value for the progressed health state was 0.53.

The new economic analyses included costs for treatment (erlotinib and pemetrexed), best supportive care, adverse events, and post-progression treatment. Costs for erlotinib were based on the patient access scheme (see section 2.3) and the treatment duration from the SATURN trial. Drug wastage was estimated based on when treatment stopped in the SATURN trial. The average per-patient drug costs for erlotinib were £7,148 for the overall stable disease population, £6,644 for patients with stable, squamous disease, and £7,976 for patients with stable, non-squamous disease. Pemetrexed costs were based on the list price to the NHS (BNF 58) and doses were based on the distribution of body surface area of patients with stable, non-squamous disease in the SATURN trial. The average per-patient drug cost for pemetrexed was £13,062. Costs for best supportive care were based on the average cost of specialist palliative care per cancer death per year reported in a publication by NICE and the University of Sheffield (2004). This is the same methodology used in other NICE technology appraisal guidance on treatments for non-small-cell lung cancer (pemetrexed for the first-line treatment of non-small-cell lung cancer and pemetrexed for the maintenance treatment of non-small-cell lung cancer). The best supportive care costs also included costs of regular monitoring (3-monthly hospital visits consisting of a consultant appointment and an outpatient CT scan).

The costs of adverse events associated with erlotinib were based on those used in NICE's previous technology appraisal guidance on erlotinib for the treatment of non-small-cell lung cancer (now replaced by NICE's technology appraisal guidance on erlotinib and gefitinib for treating non-small-cell lung cancer that has progressed after prior chemotherapy), and adjusted for inflation. The costs of adverse events associated with pemetrexed were the same as those used in NICE's technology appraisal guidance on pemetrexed for the first-line treatment of non-small-cell lung cancer. Post-progression treatment costs were based on various sources including the BNF 58 and other NICE technology appraisal guidance (pemetrexed for the first-line treatment of non-small-cell lung cancer, and pemetrexed for the maintenance treatment of non-small-cell lung cancer). Data on the number and type of post-progression treatments were collected in the SATURN trial. Because there was a lack of data on post-progression treatment from the JMEN trial, the manufacturer assumed that the costs associated with pemetrexed would be the same as those for the placebo group of the SATURN trial.

In the manufacturer's original submission, the incremental cost-effectiveness ratio (ICER) for the comparison of erlotinib with best supportive care in patients with stable disease was £47,743 per quality-adjusted life year (QALY) gained (incremental cost £7,747 and incremental benefit 0.162 QALYs). In response to the first appraisal consultation document the manufacturer revised assumptions about how much time a patient spent on treatment in the progression-free health state and the costs attributed to post-progression best supportive care, resulting in an ICER of £39,936 per QALY gained (incremental cost £7,813, incremental benefit 0.196 QALYs) for all patients with stable disease. In response to the second appraisal consultation document the manufacturer provided a revised ICER of £40,792 per QALY gained (incremental cost £7,737, incremental benefit 0.190 QALYs) for the stable disease population, using survival estimates proposed by the ERG. For patients with stable, squamous disease the ICER for erlotinib was £35,491 per QALY gained compared with best supportive care (incremental cost £7,339, incremental benefit 0.207 QALYs). In patients with stable, non-squamous disease the ICER for erlotinib was £40,020 per QALY gained compared with best supportive care (incremental cost £8,696, incremental benefit 0.218 QALYs). For the comparison of erlotinib with pemetrexed in patients with stable, non-squamous disease, the cost of erlotinib was £7,531 lower than the cost of pemetrexed. Therefore, when erlotinib was assumed to have equal or better efficacy than pemetrexed, erlotinib dominated pemetrexed (that is, it had lower costs and better efficacy). In the manufacturer's various relative efficacy scenarios, the ICER for erlotinib compared with pemetrexed ranged from £77,598 saved per QALY lost (when erlotinib was assumed to have 10% better progression-free survival and 10% worse overall survival than pemetrexed) to £511,351 saved per QALY lost (when erlotinib was assumed to have 10% worse progression-free survival and the same overall survival as pemetrexed).

The manufacturer conducted a number of deterministic sensitivity analyses. The factors that had the greatest impact on the ICERs for erlotinib compared with best supportive care were assuming that a patient spent all their time on treatment in the progression-free health state (£44,942 per QALY gained for the stable population) and using the best supportive care costs from NICE's previous technology appraisal guidance on erlotinib for the treatment of non-small-cell lung cancer (now replaced by NICE's technology appraisal guidance on erlotinib and gefitinib for treating non-small-cell lung cancer that has progressed after prior chemotherapy) rather than those from NICE's technology appraisal guidance on pemetrexed for the first-line treatment of non-small-cell lung cancer, and pemetrexed for the maintenance treatment of non-small-cell lung cancer (£44,745 per QALY gained for the stable population). The maximum ICERs for patients with stable, squamous disease and for those with stable, non-squamous disease were £40,599 per QALY gained (compared with £35,491 in the base case) and £44,589 per QALY gained (compared with £40,020 in the base case) respectively.

The ERG reviewed the clinical evidence originally submitted by the manufacturer and the additional evidence provided during consultation. It noted that the extent to which patients and investigators were truly blind to treatment allocation throughout the SATURN trial was uncertain because patients in the erlotinib group were significantly more likely to develop a rash and suffer from diarrhoea than patients in the placebo group. The ERG queried whether the results of the SATURN trial could be generalised to the UK because the trial included very few UK patients and the population was younger and fitter than would be seen in UK clinical practice. It also commented that a greater proportion of patients had first-line treatment with paclitaxel in the trial than would occur in UK clinical practice and no patients had first-line treatment that included pemetrexed, which is now the most common first-line treatment for patients with non-squamous disease (following publication of NICE's technology appraisal guidance on pemetrexed for the first-line treatment of non-small-cell lung cancer). The ERG noted inconsistencies in the reporting of post-study treatments between the manufacturer's submission and the SATURN clinical trial report. It commented that the patients in the SATURN trial received a variety of post-progression treatments, many of which are not routinely used in the UK (including pemetrexed, vinorelbine, gemcitabine and paclitaxel). Only docetaxel and erlotinib are recommended by NICE for second-line treatment of non-small-cell lung cancer. The ERG noted that this was likely to affect overall survival results. It commented that the clinical evidence for erlotinib in patients with stable disease was based on a post hoc unstratified subgroup analysis and that the evidence for patients with squamous and non-squamous disease came from a further post hoc division of this subgroup, again according to an unstratified variable. The ERG considered that the SATURN trial was not designed to perform these types of analyses and therefore the results should be interpreted with caution.

The ERG identified data on patients with stable, non-squamous disease from the JMEN trial. It was therefore able to carry out an indirect comparison of pemetrexed and erlotinib (using data from the SATURN trial) for this patient group. The results indicated a statistically significant progression-free survival benefit for pemetrexed (HR 0.64; 95% CI 0.47 to 0.89) but the difference was not statistically significant for overall survival (HR 0.93; 95% CI 0.66 to 1.30). However, the ERG considered that the results of this analysis should be interpreted with caution because of differences in the patient populations in the JMEN and SATURN studies, the limitations of using post hoc subgroup analyses, and the uncertainty about whether both studies could be generalised to patients in the UK.

The ERG reviewed the economic analyses originally submitted by the manufacturer and the additional analyses provided in response to consultation on both appraisal consultation documents. It considered that the comparison of erlotinib with best supportive care in the total population of patients with stable disease could be misleading because it could mask differences in clinical and cost effectiveness between the squamous and non-squamous disease groups. In addition, the ERG considered that results from the SATURN trial suggest that following disease progression, subsequent survival rates differ between patients with squamous and non-squamous disease, as there appears to be no difference in post-progression survival in the non-squamous disease group, whereas approximately 60% of the survival benefit associated with erlotinib maintenance treatment occurs during the post-progression period in the squamous disease group. In addition, there was statistically significant heterogeneity in some of the prognostic factors of the patients with squamous and non-squamous disease in the SATURN trial. Therefore, the ERG expressed the view that it was not appropriate to combine the histological groups and hence did not conduct any sensitivity analyses on the manufacturer's estimates for the total stable disease population.

The ERG's main criticism of the models of erlotinib compared with best supportive care in the subgroups of patients with squamous and non-squamous disease was the method used to extrapolate survival beyond the trial period. It noted that the manufacturer had not used post-progression survival data directly from the trial, but instead calculated post-progression survival as the difference between overall survival and progression-free survival. The ERG commented that because all patients in the stable disease population of the SATURN trial had disease which had progressed (that is, the progression-free survival data set was complete), there was no need to model the mean duration of progression-free survival because it could be based directly on Kaplan-Meier survival estimates from the trial. The ERG estimated mean overall survival using new survival data provided by the manufacturer during consultation on the first appraisal consultation document, by adding progression-free survival to post-progression survival and adjusting the estimate to exclude patients dying at or before disease progression. For post-progression survival in the subgroup of patients with squamous disease, the ERG used Kaplan-Meier estimates up to a 20% survival figure and then modelled the remaining survival. For the subgroup of patients with non-squamous disease, data on post-progression survival were available up to 600 days, but required modelling after this point for the remaining 7% of patients; a common exponential model was used for both the intervention and the comparator. The ERG's approach to modelling survival resulted in mean overall survival gains of 3.4 months (95% CI 1.5 to 5.3 months) for patients with stable, squamous disease and 2.2 months (95% CI 0.9 to 3.5 months) for patients with stable, non-squamous disease (corresponding to a 28% and 75% decrease in QALY gains compared with those in the manufacturer's base case respectively). The ERG stressed the wide confidence intervals around these estimates of survival gain. For the population of patients with stable, squamous disease the manufacturer's base-case ICER for erlotinib compared with best supportive care increased from £35,491 to £44,812 per QALY gained (incremental cost £7,129, incremental benefit 0.1591 QALYs) using the ERG's modelling approach. For the population of patients with stable, non-squamous disease the manufacturer's base-case ICER for erlotinib compared with best supportive care increased from £40,020 to £68,120 per QALY gained (incremental cost £8,340, incremental benefit 0.1224 QALYs).

The ERG commented on the manufacturer's economic analysis of erlotinib compared with pemetrexed in the population of patients with stable, non-squamous disease. It noted that the manufacturer's estimate of pemetrexed costs did not account for gender differences in body surface area. It also noted that a modifying factor that reduced pemetrexed costs by 5% had been used and that no evidence had been provided to support this method. The ERG revised the manufacturer's model using their own estimation of pemetrexed costs, and used the hazard ratios from its indirect analysis (noting the previously mentioned caution about this analysis). The resulting ICER was based on the lower costs (-£8,460) and lower efficacy (-0.1007 QALYs) of erlotinib compared with pemetrexed and represented a cost saving of £84,029 per QALY lost.

Full details of all the evidence are in the manufacturer's submission and the ERG report.